Release version 6.0.0

Two defects kept dedicated Postgres databases out of the backup.

docker exec never forwarded PGPASSWORD. execute_to_file set it on baudolo's
own process, but nothing carried it across the container boundary, so an
engine whose pg_hba demands a password on TCP loopback refused the dump.
forward_env passes a bare `-e NAME`, letting docker copy the value out of
this process's environment instead of spelling it into argv, where the
host's process list would publish it.

get_instance returned the container name unchanged when that name carried no
database token, claiming an instance it had never derived. An application
container therefore answered the same databases.csv row as its own dedicated
engine, and application images often ship the engine's client tools, so the
dump command started and wrote a file that looked like a backup and held
none of the data. Discourse is the live case: its launcher names the
container `discourse`, and the image ships pg_dumpall.

The regex stays a normaliser - `<app>-database` from compose and
`<app>_database.1.<task>` from swarm still resolve to the same instance.
Only the fallthrough changes.

BREAKING CHANGE: a database container whose name carries no `database`, `db`
or `postgres` token must now be named in --database-containers. Without that
declaration its rows no longer match and no dump is written.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.claude/.gitignore

@@ -0,0 +1,3 @@
+*
+!.gitignore
+!settings.json

.claude/settings.json

@@ -0,0 +1,11 @@
+{
+  "permissions": {
+    "ask": [
+      "Bash(git commit*)",
+      "Edit(CHANGELOG.md)",
+      "Write(CHANGELOG.md)",
+      "Edit(pyproject.toml)",
+      "Write(pyproject.toml)"
+    ]
+  }
+}

.dockerignore

@@ -0,0 +1,13 @@
+.git
+.github
+__pycache__
+**/__pycache__
+*.egg-info
+**/*.egg-info
+artifacts/
+dist/
+build/
+.venv
+.ruff_cache
+.pytest_cache
+.mcp.json

.github/dependabot.yml

@@ -0,0 +1,21 @@
+---
+version: 2
+updates:
+  - package-ecosystem: github-actions
+    directory: /
+    schedule:
+      interval: weekly
+    groups:
+      actions:
+        patterns:
+          - "*"
+
+  - package-ecosystem: docker
+    directory: /
+    schedule:
+      interval: weekly
+
+  - package-ecosystem: pip
+    directory: /
+    schedule:
+      interval: weekly

.github/workflows/ci.yml

@@ -22,20 +22,29 @@ jobs:
 
     steps:
       - name: Checkout
-        uses: actions/checkout@v4
+        uses: actions/checkout@v7
 
       - name: Show docker info
         run: |
           docker version
           docker info
 
+      - name: Provide zfs so the snapshot suite covers every filesystem
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends zfsutils-linux
+          sudo modprobe zfs
+          zpool version
+
       - name: Run all tests via Makefile
+        env:
+          E2E_REQUIRE_FILESYSTEMS: "btrfs ext4 zfs"
         run: |
           make test
 
       - name: Upload E2E artifacts (always)
         if: always()
-        uses: actions/upload-artifact@v4
+        uses: actions/upload-artifact@v7
         with:
           name: e2e-artifacts
           path: artifacts
@@ -49,7 +58,7 @@ jobs:
 
     steps:
       - name: Checkout (full history for tags)
-        uses: actions/checkout@v4
+        uses: actions/checkout@v7
         with:
           fetch-depth: 0
 
@@ -67,7 +76,7 @@ jobs:
           git push -f origin stable
 
       - name: Login to GHCR
-        uses: docker/login-action@v3
+        uses: docker/login-action@v4
         with:
           registry: ${{ env.REGISTRY }}
           username: ${{ github.actor }}

.github/workflows/dependabot-auto-merge.yml

@@ -0,0 +1,29 @@
+---
+name: Dependabot auto-merge
+
+on: pull_request
+
+permissions:
+  contents: write
+  pull-requests: write
+
+jobs:
+  auto-merge:
+    # Enable auto-merge for minor/patch dependency bumps; majors stay manual.
+    # The merge only happens once every required status check (make test) is
+    # green, so the CI stays the gate.
+    if: github.actor == 'dependabot[bot]'
+    runs-on: ubuntu-latest
+    steps:
+      - name: Fetch dependency metadata
+        id: metadata
+        uses: dependabot/fetch-metadata@v3
+        with:
+          github-token: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Enable auto-merge (minor + patch)
+        if: steps.metadata.outputs.update-type != 'version-update:semver-major'
+        run: gh pr merge --auto --squash "$PR_URL"
+        env:
+          PR_URL: ${{ github.event.pull_request.html_url }}
+          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}

.gitignore

@@ -1,2 +1,6 @@
 __pycache__
-artifacts/
+artifacts/
+*.egg-info
+dist/
+build/
+.mcp.json

CHANGELOG.md

@@ -1,3 +1,538 @@
+# Changelog
+
+## [6.0.0] - 2026-08-18
+
+Breaking:
+- Backup: a database container whose name carries no *database*, *db* or
+  *postgres* token — preceded by a hyphen or underscore — must now be named in
+  *--database-containers*. Without that declaration its *databases.csv* rows no
+  longer match and no dump is written, silently, because nothing fails. The
+  shape this hits hardest is a container named exactly *postgres* or *mariadb*:
+  the token needs a separator in front of it, which a bare name does not have.
+  *app-database*, *app_database.1.<task>* from swarm and *app-postgres-1* are
+  unaffected, as is any container already declared.
+
+Fixed:
+- Backup: *docker exec* now forwards *PGPASSWORD* into the container.
+  *execute_to_file* set the variable on baudolo's own process, but nothing
+  carried it across the container boundary, so an engine whose *pg_hba* demands
+  a password on TCP loopback refused every dump — which is every dedicated
+  Postgres instance on a real host. The name travels as a bare *-e NAME* so
+  docker copies the value out of this process's environment; spelling
+  *-e NAME=value* instead would publish the secret in the host's process list.
+- Backup: *get_instance* no longer claims an instance it never derived. It
+  returned the container name unchanged when that name carried no database
+  token, so an application container answered the same *databases.csv* row as
+  its own dedicated engine. Application images frequently ship the engine's
+  client tools, so the dump command started and wrote a file that looked like a
+  backup and held none of the data: measured against Discourse, 1,680 bytes
+  from the application where the engine produced 10,469,439. The regex stays a
+  normaliser — *<app>-database* from compose and *<app>_database.1.<task>* from
+  swarm still resolve to one instance. Only the fallthrough changed.
+
+New:
+- Tests: *get_instance* has unit coverage for the first time. Eleven cases pin
+  the container names that compose, swarm and explicitly-named engines produce,
+  so a future change to the regex has to state which shape it gives up.
+- Tests: two e2e modules cover shapes the suite structurally could not see.
+  Every fixture passed its container in *--database-containers*, which left the
+  regex branch — the only one a dedicated database ever takes — dead code under
+  test, and no scenario made a password mandatory, because stock
+  *postgres:alpine* grants trust on loopback.
+  *test_e2e_postgres_password_required* starts an engine with
+  *--auth-host=scram-sha-256* and carries a negative control asserting the
+  server refuses an unauthenticated dump; without it the module would pass
+  whether or not the password is forwarded at all.
+  *test_e2e_app_container_ships_client_tools* places an application container
+  beside its engine with neither declared, and requires the engine dumped, the
+  application volume copied as files, and no dump written from the application.
+
+## [5.0.0] - 2026-08-18
+
+**[5.0.0] - 2026-08-18**
+
+Breaking:
+- Library: *BackupException* is now *BackupError* and *VersionMismatch* is now
+  *VersionMismatchError*. Both names violated the convention that an exception
+  class ends in *Error*; the first is imported by six modules, so the rename is
+  atomic across the package.
+- Library: *backup_dumps_for_volume* and *backup_mariadb_or_postgres* return a
+  *VolumeOutcome* instead of a *(bool, bool)* tuple. The pair could not carry
+  the detected engine, which the caller needs for the manifest.
+
+New:
+- Backup: every generation carries a *manifest.json* stating its layout and,
+  per volume, *database* (it held one), *dumped* (a dump was produced) and
+  *engine* (which one was detected). A volume with *database* and no *dumped*
+  was copied as raw engine files — under *--only-sql* that fallback is the
+  documented behaviour, and until now nothing in the finished tree said it had
+  happened. Restoring such a volume replays engine files instead of a dump.
+- Library: *baudolo.generation* states the generation layout once — *files*,
+  *sql*, the dump suffixes, the manifest name. *BackupPaths*, the dump writer
+  and the volume copier stop spelling them out separately. The module is
+  import-free on purpose, so a consumer can read a manifest with nothing but
+  *json* on a host where this package is not installed.
+
+Changed:
+- Build: the test targets no longer depend on *clean*. *clean* is
+  *git clean -fdX .*, so running the unit tests deleted every git-ignored file
+  in the working tree. It was compensating for a missing *.dockerignore*, which
+  now keeps *__pycache__*, egg-info and build output out of the image context
+  where that belongs. *clean* remains available as its own target.
+- Lint: a ruff configuration is declared. The package ran on ruff's defaults
+  while its consumer held itself to a far wider selection; measured against
+  that selection the tree had 224 findings and now has none. Includes a full
+  *os.path* to *pathlib* migration, with two deliberate exceptions: *abspath*
+  stays where *Path.resolve()* would follow symlinks and let a symlinked volume
+  test as inside the snapshot subject, and the rsync trailing separator is kept
+  explicit where *Path* would drop it.
+
+## [4.0.0] - 2026-08-17
+
+Breaking:
+- CLI: *--repo-name* and *--databases-csv* are required. One default was the
+  literal *backup-docker-to-local* while its help promised the git folder name;
+  the other pointed into the installed package, so a forgotten flag ran the
+  whole backup silently without a single dump.
+- CLI: *--dump-only-sql* is now *--only-sql*; the old spelling exits 2.
+- CLI: *--everything* is withdrawn. Its only real effect was to ignore
+  *--images-no-stop-required*, which leaving that list empty already does.
+- Library: the runner injected into *volume_snapshot* receives an argv list
+  instead of a command string.
+
+New:
+- Backup: *--only-files* — no dumps at all, every volume as files, for hosts
+  that hold no database credentials. Needs no *--databases-csv*; mutually
+  exclusive with *--only-sql*.
+- Backup: the engine is detected by executing the dump tool in the container
+  (*pg_dumpall*, *mariadb-dump*, *mysqldump*), not by reading the image name.
+  A dedicated Postgres tagged *<app>-database* is finally dumped; an image
+  merely named like an engine no longer kills the run with exit 127. Probed
+  once per image ID, and an image shipping only *mysqldump* is dumped with it.
+- Library: *baudolo.databases* states the databases.csv contract once —
+  columns, delimiter, cluster marker, validator, *read_rows()* — for the seed,
+  the backup, and external consumers.
+
+Changed:
+- Backup: every command is an argv list; *shell=True* is gone. A database name
+  is validated on read as strictly as the seed writes it, *PGPASSWORD* travels
+  in the child's environment instead of the command string, and a failing dump
+  deletes its partial file instead of leaving it behind.
+
+## [3.6.1] - 2026-08-17
+
+- Restore: *--empty* on a cluster dump is a catalog-wide sweep — it drops every
+  non-template database and every non-pg_ role of the instance. On a dedicated
+  instance that is right, because the dump recreates all of it; on a shared one
+  it destroys databases the dump does not carry, with nothing to restore them
+  from. No test had ever executed that sweep: the e2e dropped the cluster by
+  hand first and left the pre-clean with zero rows to generate.
+- Restore: the instance is checked instead of the sweep being narrowed.
+  *--empty* refuses when the instance holds a database the dump does not carry,
+  names it, and touches nothing. Narrowing the sweep is the obvious fix and is
+  worse — a surviving database that owns or merely grants to one of the dump's
+  roles pins it in *pg_shdepend*, *DROP OWNED BY* reaches only the control
+  database the pre-clean is connected to, so *DROP ROLE* fails after the dump's
+  own databases are already gone and the replay never starts.
+- Restore: the dump's inventory is read with a real identifier parser. A quoted
+  name may hold spaces, and psql options precede the target of a *\connect*
+  line, so a character class that stops at whitespace read *"odd name"* as
+  *odd* and *-reuse-previous=on* as a database.
+- Tests: the cluster e2e no longer empties the instance itself, so *--empty*
+  has to do it and the replay has to put it back. A second pass adds a foreign
+  database and requires the refusal to leave both it and the restored data
+  untouched.
+
+## [3.6.0] - 2026-08-17
+
+- Restore: *--empty* drops the schema in one session and replays in the next,
+  with no rollback across the two, so a dump the engine could not parse left an
+  emptied database behind. The dump's header is now checked against the running
+  engine before anything is dropped, and a newer dump is refused.
+  Forward across a major version stays allowed; *--no-version-check* is the way out.
+- Restore: a volume with driver options — NFS, a bind device, tmpfs — keeps the
+  usual *_data* path, but docker mounts its real storage over it only while a
+  container holds it. Restoring meanwhile landed under the mount, stayed hidden
+  there, and rsync reported success. That volume is now refused until something
+  mounts it.
+- Backup: the same volume sits in a snapshot as an empty directory, so it was
+  copied empty and the generation stamped complete. Capture is decided per volume
+  now — an uncaptured one is copied live, the rest keep their snapshot. A single
+  NFS volume no longer costs the whole host its consistent backup.
+
+## [3.5.0] - 2026-08-17
+
+- Restore: a *database = '*'* row makes the backup write
+  *<instance>.cluster.backup.sql* via *pg_dumpall*, and nothing could read it
+  back — the CLI knew *files*, *postgres* and *mariadb*, so that dump was
+  stored and unrestorable. *baudolo-restore cluster* replays it against the
+  control database, deliberately without *--single-transaction* because
+  CREATE DATABASE is forbidden inside a transaction block, and filters out the
+  CREATE ROLE of the connecting role, which the pre-clean cannot drop while it
+  holds the session. *--empty* drops the cluster's databases first, then
+  releases what its roles still own, then the roles themselves.
+
+- Lint: ruff was never wired into the repository — no target, no CI step, no
+  pin — and reported 45 findings across sources and tests. *make ruff* and
+  *make lint* now run it over every file, *make test* gates on a clean run as a
+  fourth parallel spur, and the linter is pinned in a *lint* extra because a
+  minor bump changes which rules fire.
+
+## [3.4.3] - 2026-08-16
+
+- Backup: *create_version_directory* carried *exist_ok=True*, so a run starting
+  in the same wall-clock second as its predecessor claimed that predecessor's
+  generation. Generation names carry seconds, and a host with little to copy
+  finishes inside one — rsync *--delete* then overwrote a finished generation,
+  and only afterwards did *create_stamp_file* refuse the already-stamped
+  directory and exit 2. The guard reported the damage instead of preventing it.
+- Backup: the generation directory is claimed exclusively. Claiming it is the
+  first filesystem action of a run, so a collision aborts before the first
+  write and names the second it collided on.
+- Tests: the idempotence test asserted the reuse and gave way to one that
+  requires the refusal.
+
+## [3.4.2] - 2026-08-15
+
+- Backup: *has_image* matched the raw *.Config.Image*, so the registry host and
+  the tag decided the dump tool. On a swarm node named after the app under test,
+  *svc-db-mariadb-swarm-mgr-01:5000/postgres_custom* read as MariaDB and
+  *mariadb-dump* ran inside a Postgres container: exit 127, and the
+  *BackupException* took the backup unit with it.
+- Backup: *image_name* strips digest, tag and registry host, so the engine rests
+  on the repository path alone — the exact-matching intent of 3.0.0 applied to
+  the one place that change did not reach.
+- Tests: both false-positive directions on *has_image*, plus an e2e that
+  reproduces the shape with a *docker tag* and asserts a real *pg_dump* lands.
+
+## [3.4.1] - 2026-08-05
+
+- Backup: each volume is copied twice into the same destination — once hot,
+  once cold after the container is stopped — and rsync ran with *-b*, so
+  *--delete* renamed rather than removed a file the source had dropped between
+  the passes. Stopping a container is what makes the source drop files: a
+  graceful shutdown flushes and the format rolls its commit point. The
+  superseded file survived as *name~* beside the real one and was restored into
+  live data.
+- Backup: for an opaque payload that is stale bytes nobody reads; for a format
+  that enumerates its own directory it is corruption. Lucene resolves the
+  current commit by parsing every file starting with *segments* as a radix-36
+  generation, so a restored *segments_3~* leaves the shard store unreadable and
+  the primary at *NO_VALID_SHARD_COPY*. With *.security-7* unallocatable the
+  reserved *elastic* user has no password hash, every probe answers 401 and the
+  container never turns healthy.
+- Backup: *--link-dest* already provides the incrementals and nothing reads the
+  twins — the restore path is an unfiltered *rsync -avv --delete* into the live
+  volume. Dropping *-b* leaves the predecessor generation byte-identical, keeps
+  the hardlinks intact and makes generations smaller, never larger.
+- Tests: the absence of *--backup* is asserted on the rsync invocation.
+
+## [3.4.0] - 2026-08-02
+
+- Backup: *-a* implies *-D*, so a generation was written with
+  *--devices --specials* and rsync recreated every unix socket and fifo found in
+  a volume. Where the backup root is an nfs-ganesha export, ganesha accepts the
+  socket on write but cannot serve it back, and the remote pull's sender then
+  fails with *readdir* / *readlink_stat* "Invalid argument (22)" and exits 23 —
+  deterministically, for every retry. *--no-D* keeps them out of the generation.
+- Backup: nothing restorable is lost. Sockets and fifos are recreated by the
+  daemons that own them, and the postfix queue itself — *incoming*, *active*,
+  *deferred*, *hold*, *maildrop* — is unaffected, so accepted-but-undelivered
+  mail stays in the backup. Device nodes go too; the only volume that could hold
+  them is a nested docker data root, which does not belong in a backup anyway.
+- Tests: the flag is asserted on the rsync invocation.
+
+## [3.3.0] - 2026-08-02
+
+- Backup: *--volumes-no-backup-required* excludes a volume by name.
+  *--images-no-backup-required* resolves through *volume_is_fully_ignored*,
+  which skips a volume only when every container using it is ignored — a
+  container holding a derived tree beside state that must be kept cannot
+  express the exclusion at all. A docker-in-docker data root is exactly that
+  shape, and excluding by image would drop all three of its volumes.
+- Backup: the name check runs before *containers_using_volume*, so an excluded
+  volume costs no docker inspection and the decision does not depend on which
+  containers exist when the run starts.
+- Tests: two volumes off one container, asserting the sibling survives — the
+  property the image lever cannot provide — as unit and end-to-end.
+
+## [3.2.2] - 2026-07-31
+
+- Backup: the btrfs snapshot is carved inside its subject, as
+  *<data root>/.baudolo-<tag>*, not beside it. The kernel refuses a snapshot
+  whose destination is on another filesystem, which is exactly what the parent
+  directory is when the data root is a mountpoint of its own — a dedicated disk
+  mounted onto */var/lib/docker* failed every run with EXDEV. Placing it inside
+  makes source and destination the same filesystem by construction, and aligns
+  btrfs with the zfs path, which already resolves its snapshot inside the
+  subject at *<subject>/.zfs/snapshot/<tag>*. A leftover from an interrupted run
+  appears in the next snapshot as an empty directory rather than recursing,
+  since btrfs does not include nested subvolumes.
+
+## [3.2.1] - 2026-07-31
+
+- Backup: the snapshot resolver keeps the trailing separator *get_storage_path*
+  puts on a volume path — *os.path.abspath* stripped it. rsync reads *dir* as
+  "copy the directory" where *dir/* means "copy its contents", so every snapshot
+  generation landed at *<volume>/files/_data/...* while the live path lands at
+  *<volume>/files/...*. Restores read the live layout, and *--link-dest* had
+  nothing to match against the previous generation.
+- Backup: snapshot teardown no longer fails a completed run. A busy
+  *btrfs subvolume delete* raised out of the *finally*, skipping the generation
+  stamp and the compose handling on a run whose data was already copied, and
+  masking whatever the body had raised. The leftover is reported instead.
+- Backup: a volume created after the snapshot was taken is copied live with a
+  warning instead of aborting the run. Nothing is stopped in snapshot mode, so
+  the host keeps creating volumes for the whole copy.
+- Backup: the snapshot pass compares by content (*--checksum*) again. 3.2.0
+  dropped it because a snapshot source cannot move, which is true, but the
+  comparison that matters is against *--link-dest*: a file that changed while
+  keeping its size and whole-second mtime was hard-linked stale out of the
+  previous generation, and the single pass had no authoritative pass to repair
+  it. Still one pass where the live path takes two.
+- Backup: *--hard-restart-projects* is refused alongside *--snapshot*, like
+  *--shutdown* already is. It exists for stacks whose database cannot be backed
+  up hot, which is what a snapshot removes.
+- Tests: the trailing separator, both teardown behaviours, the new refusal, and
+  *app.main* driving the snapshot branch — the caller that runs in production,
+  which no test had exercised, which is why the layout defect shipped.
+
+## [3.2.0] - 2026-07-31
+
+- Backup: *--snapshot {btrfs,zfs}* with *--snapshot-subject* captures every
+  volume from one atomic filesystem snapshot. The subject — the btrfs subvolume
+  or zfs dataset holding the docker volumes, e.g. */var/lib/docker* — is frozen
+  once per run, so a generation shares a single point in time and no container
+  is stopped. Restoring such a copy is an ordinary crash recovery, which every
+  supported engine performs at startup. This is the mode 3.1.4 pointed to for
+  volumes where two rsync passes over a live tree stop being affordable.
+- Backup: snapshot passes copy once and drop *--checksum*. Verification exists
+  because a hot pass writes its destination from a moving source; a snapshot
+  source cannot move, so size and mtime cannot race and the second full read is
+  pure cost.
+- Backup: an unsupported filesystem or unknown kind fails with *SnapshotError*.
+  The kind is stated, not probed — an inconclusive probe would fall back to a
+  live copy and hand out the torn backup the mode prevents. *--shutdown* is
+  rejected alongside *--snapshot* rather than ignored, since nothing is stopped.
+- Refactor: *backup/app.py* splits into *layout.py*, *policy.py* and *dumps.py*
+  along the lines it already had; 276 lines down to 124.
+- Tests: unit coverage for snapshot, layout, policy, volume and CLI validation.
+  E2E cases drive real btrfs, zfs and ext4 on loop devices, one proving the
+  loud refusal; another writes a MariaDB across the snapshot and requires the
+  restored server to recover on its own with every committed row and none of
+  the later ones. CI installs zfs and sets *E2E_REQUIRE_FILESYSTEMS*, so a
+  missing kernel module fails the build instead of skipping a filesystem.
+
+## [3.1.4] - 2026-07-31
+
+- Backup: the post-stop volume pass now compares by content (*--checksum*), so
+  it can no longer skip a file the hot pass had copied from a live source. Each
+  volume is rsynced twice into the same destination — once with the container
+  running, once after it is stopped — and the second pass used rsync's quick
+  check (size plus whole-second mtime). A pre-allocated 16 MiB WAL segment never
+  changes size, so when its last pre-stop write and postgres' shutdown
+  checkpoint fell in the same whole second, the cold pass skipped it while still
+  replacing *global/pg_control*, whose previous write was seconds earlier. The
+  generation then paired a post-shutdown *pg_control* with a WAL segment still
+  zeroed at the recorded checkpoint LSN, and restoring it crash-looped postgres
+  with "invalid record length … expected at least 24, got 0" followed by "PANIC:
+  could not locate a valid checkpoint record". *backup_volume* takes
+  *authoritative* as a required keyword rather than an optional flag, so each of
+  the four call sites states which pass it is; the hot passes keep the quick
+  check. *--ignore-times* was rejected because it destroys the *--link-dest*
+  hardlink dedup (measured 20/20 to 0/20, generation size doubled), and
+  *--modify-window=-1* because it makes correctness depend on the destination
+  filesystem preserving sub-second mtimes, which degrades silently on NFS or
+  ext3.
+- Cost: the quick check scales with file count, *--checksum* with bytes read on
+  both sides, and it runs while the container is stopped. The extra stop time
+  stays under a minute up to roughly 4 GB on spinning disk, 15 GB on a SATA SSD
+  and 60 GB on NVMe; at 1 TB it is 17 minutes on NVMe and over three hours on
+  spinning disk. No size threshold is built in, since a guessed one would drop
+  the guarantee exactly where an unrestorable backup costs most — volumes at
+  that scale want filesystem snapshots or *pg_basebackup* rather than two rsync
+  passes over a live tree.
+
+## [3.1.3] - 2026-07-20
+
+- Restore: the postgres dump replay now runs under *--single-transaction*,
+  so a concurrent writer on a live database can no longer interleave a row
+  between the replay's table re-create and its *COPY* and trip a "duplicate
+  key value violates unique constraint" abort under ON_ERROR_STOP. This is
+  the discourse restore-drill race (*mini_scheduler* upserting
+  *scheduler_stats(id=1)* mid-restore) that failed the whole restore. The
+  *--empty* pre-clean stays multi-statement (*\gexec*, one DROP per
+  statement) because a single DROP transaction exhausts
+  *max_locks_per_transaction* on large schemas (e.g. gitlab).
+- Refactor: the *--empty* pre-clean SQL moves out of the inline Python
+  string into *src/baudolo/restore/db/empty_preclean.sql* (loaded via
+  *dirname(__file__)*, declared as package-data so it ships in the wheel).
+- Tests: a unit test guards the single-transaction / multi-statement split
+  (replay carries *--single-transaction*, pre-clean does not); a new e2e
+  reproduces the live-writer race and asserts the restore survives it.
+
+## [3.1.2] - 2026-07-18
+
+- Restore: the postgres *--empty* pre-clean also drops user-owned text
+  search configurations and dictionaries (*pg_ts_config*, *pg_ts_dict*),
+  so a schema shipping a custom dictionary (e.g. taiga's
+  *english_stem_nostop*) no longer aborts the replay with "duplicate key
+  value violates unique constraint pg_ts_dict_dictname_index" under
+  ON_ERROR_STOP.
+- Tests: the string-assertion unit test for the pre-clean SQL is replaced
+  by real scenario data in the e2e: the seeded schema contains an
+  overloaded *f()/f(int)* pair and the nostop dictionary plus
+  configuration, and the restored database is queried to prove each
+  survives the backup, pre-clean and replay cycle exactly once.
+
+## [3.1.1] - 2026-07-17
+
+- Restore: the postgres *--empty* pre-clean drops functions and procedures
+  by their identity signature (*pg_get_function_identity_arguments*), so a
+  schema that overloads a function name (e.g. discourse) no longer aborts
+  the replay with "function name is not unique" under ON_ERROR_STOP.
+  Identifier quoting moves from the outer DROP format into each object
+  branch, since the *name(args)* compound must not be quoted as a whole; a
+  unit test pins the per-branch *%I* quoting so future branches cannot
+  regress unquoted.
+
+## [3.1.0] - 2026-07-15
+
+- Restore: the postgres *--empty* pre-clean emits one DROP per object and
+  runs them via *\gexec* instead of a single DO-block, so large schemas
+  (e.g. gitlab) no longer exhaust *max_locks_per_transaction* in one
+  transaction. It also drops user-owned non-public schemas, so dumps that
+  CREATE SCHEMA (e.g. discourse's *discourse_functions*) no longer abort
+  on the already-existing schema under ON_ERROR_STOP.
+- Backup: *--database-containers* and *--images-no-stop-required* are now
+  optional and default to an empty list, so a pure file backup needs no
+  dummy arguments; an empty stop whitelist keeps the conservative
+  stop-all behavior.
+- Tests: new e2e test restores *--empty* against a fully populated
+  database containing a non-public schema and every dropped object class.
+  *make test* runs the three suites concurrently after a single
+  clean+build; *E2E_TEST_PATTERN* runs an e2e subset.
+
+## [3.0.0] - 2026-07-12
+
+- Backup: *--images-no-stop-required* and *--images-no-backup-required* now
+  match a container's exact *.Config.Image* (full *repo:tag*, registry
+  prefix included) instead of a substring, so a near-miss image name no
+  longer flips the stop/skip decision. Callers must pass exact image
+  references. **Breaking.**
+- Backup: renamed *--hard-compose-restart* to *--hard-restart-projects*
+  (its value stays a list of compose project dir names). **Breaking:** the
+  old flag name is removed.
+
+## [2.0.0] - 2026-07-12
+
+- Backup: renamed *--docker-compose-hard-restart-required* to
+  *--hard-compose-restart* and changed its default from *["mailu"]* to *[]*
+  (nargs="*"). The compose down/up is now opt-in: compose hosts pass
+  *mailu* explicitly, while swarm hosts pass nothing, since there the dir is
+  a stack whose overlay network collides with *compose up*. **Breaking:** the
+  old flag name is removed and the implicit mailu default is gone.
+- Backup: *--backups-dir* is now required (no */var/lib/backup/* default) so
+  a run can never silently target the wrong backup root. **Breaking.**
+- Restore: volume files are rsynced directly into the target volume's
+  mountpoint (resolved via *docker volume inspect*), mirroring the backup
+  path; the *alpine-rsync* helper image and the *--rsync-image* flag are
+  gone. The caller needs write access to the docker volume root (root on the
+  host, baudolo's normal privilege). **Breaking:** the restore *files*
+  subcommand no longer accepts *--rsync-image*.
+- Tests: the e2e suite tracks *postgres:alpine* (18+, mounted at
+  */var/lib/postgresql*) and *mariadb:latest* from a single source of truth.
+
+## [1.8.1] - 2026-07-12
+
+- Restore: the postgres empty mode also drops user-owned collations in
+  public; dumps containing CREATE COLLATION (e.g. OpenProject's ICU
+  collation versions_name) no longer abort the replay with 'collation
+  already exists'.
+- Maintenance: base image bumped from python 3.11-slim to 3.14-slim.
+
+## [1.8.0] - 2026-07-11
+
+Swarm-aware backups and replayable restores.
+
+- Backup: swarm task containers are never stopped or started manually
+  anymore; they are skipped visibly and backed up hot, while the sql dump
+  stays the consistent database backup.
+- Backup: a container that vanishes between listing and inspect no longer
+  aborts the run; a failing inspect on a container that still exists keeps
+  failing loudly.
+- Backup: pg_dump runs with the no-owner and no-privileges flags so dumps
+  are replayable by the owning app user.
+- Restore: the mariadb empty mode drops all tables in one client session
+  with FOREIGN_KEY_CHECKS disabled; FK-linked parent tables no longer abort
+  the replay with ERROR 1451.
+- Restore: the postgres empty mode drops only current-user-owned objects,
+  and the replay skips superuser-only dump lines without ever touching
+  COPY data blocks.
+- Restore: the replay streams the dump through a temp file instead of
+  buffering it in memory; multi-GB dumps no longer OOM the restore.
+- Tooling: the e2e runner reaches the DinD daemon via docker exec instead
+  of a host-published unencrypted API port.
+- Tooling: new end-to-end test reproducing the swarm stop flake, plus unit
+  tests for the restore filters and the swarm probes; the suite is 36 unit,
+  9 integration and 30 e2e tests.
+- Tooling: Dependabot with auto-merge for minor and patch updates.
+
+## [1.7.1] - 2026-05-26
+
+* 🔌 MariaDB SQL backups now connect over TCP loopback so the dump always matches the same wildcard-host grant the application uses — no more surprise `ERROR 1045 Access denied` when a localhost-bound auth row preempts.
+* 🧪 New regression and bug-repro tests pin the TCP behaviour and prove it under the exact preemption setup that caused the production failure on MariaDB 12.
+* 🩺 E2E test infrastructure: DinD bridge and inner daemon now default to MTU 1280 so registry pulls survive host paths with broken PMTUD (override via `E2E_DIND_MTU`).
+
+
+## [1.7.0] - 2026-02-07
+
+* 🚀 Backup jobs now support all valid Docker Compose file names – case-insensitive and hassle-free.
+
+
+## [1.6.0] - 2026-02-06
+
+* Compose handling is now fully delegated to the Infinito.Nexus compose wrapper or plain docker compose, removing all custom env and file detection to ensure a single, consistent source of truth.
+
+
+## [1.5.0] - 2026-01-31
+
+* * Make `databases.csv` optional: missing or empty files now emit warnings and no longer break backups
+* Fix Docker CLI compatibility by switching to `docker-ce-cli` and required build tools
+
+
+## [1.4.0] - 2026-01-31
+
+* Baudolo now restarts Docker Compose stacks in a wrapper-aware way (with a `docker compose` fallback), ensuring that all Compose overrides and env files are applied identically to the Infinito.Nexus workflow.
+
+
+## [1.3.0] - 2026-01-10
+
+* Empty databases.csv no longer causes baudolo-seed to fail
+
+
+## [1.2.0] - 2025-12-29
+
+* * Introduced **`--dump-only-sql`** mode for reliable, SQL-only database backups (replaces `--dump-only`).
+* Database configuration in `databases.csv` is now **strict and explicit** (`*` or concrete database name only).
+* **PostgreSQL cluster backups** are supported via `*`.
+* SQL dumps are written **atomically** to avoid corrupted or empty files.
+* Backups are **smarter and faster**: ignored volumes are skipped early, file backups run only when needed.
+* Improved reliability through expanded end-to-end tests and safer defaults.
+
+
+## [1.1.1] - 2025-12-28
+
+* * **Backup:** In ***--dump-only-sql*** mode, fall back to file backups with a warning when no database dump can be produced (e.g. missing `databases.csv` entry).
+
+
+## [1.1.0] - 2025-12-28
+
+* * **Backup:** Log a warning and skip database dumps when no databases.csv entry is present instead of raising an exception; introduce module-level logging and apply formatting cleanups across backup/restore code and tests.
+* **CLI:** Switch to an FHS-compliant default backup directory (/var/lib/backup) and use a stable default repository name instead of dynamic detection.
+* **Maintenance:** Update mirror configuration and ignore generated .egg-info files.
+
+
 ## [1.0.0] - 2025-12-27
 
 * Official Release 🥳

Dockerfile

@@ -1,34 +1,37 @@
 # syntax=docker/dockerfile:1
-FROM python:3.11-slim
+FROM python:3.14-slim
 
 WORKDIR /app
 
-# Runtime + build essentials:
-# - rsync: required for file backup/restore
-# - ca-certificates: TLS
-# - docker-cli: needed if you want to control the host Docker engine (via /var/run/docker.sock mount)
-# - make: to delegate install logic to Makefile
-#
-# Notes:
-# - On Debian slim, the docker client package is typically "docker.io".
-# - If you only want restore-without-docker, you can drop docker.io later.
+# Base deps for build/runtime + docker repo key
 RUN apt-get update && apt-get install -y --no-install-recommends \
     make \
     rsync \
     ca-certificates \
-    docker-cli \
+    bash \
+    curl \
+    gnupg \
   && rm -rf /var/lib/apt/lists/*
 
+# Install Docker CLI (docker-ce-cli) from Docker's official apt repo
+RUN bash -lc "set -euo pipefail \
+  && install -m 0755 -d /etc/apt/keyrings \
+  && curl -fsSL https://download.docker.com/linux/debian/gpg \
+     | gpg --dearmor -o /etc/apt/keyrings/docker.gpg \
+  && chmod a+r /etc/apt/keyrings/docker.gpg \
+  && . /etc/os-release \
+  && echo \"deb [arch=\$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.gpg] https://download.docker.com/linux/debian \${VERSION_CODENAME} stable\" \
+     > /etc/apt/sources.list.d/docker.list \
+  && apt-get update \
+  && apt-get install -y --no-install-recommends docker-ce-cli \
+  && rm -rf /var/lib/apt/lists/*"
+
 # Fail fast if docker client is missing
+RUN docker version || true
 RUN command -v docker
 
 COPY . .
-
-# All install decisions are handled by the Makefile.
 RUN make install
 
-# Sensible defaults (can be overridden at runtime)
 ENV PYTHONUNBUFFERED=1
-
-# Default: show CLI help
 CMD ["baudolo", "--help"]

MIRRORS

@@ -1,4 +1,4 @@
 git@github.com:kevinveenbirkenbach/backup-docker-to-local.git
 ssh://git@git.veen.world:2201/kevinveenbirkenbach/backup-docker-to-local.git
 ssh://git@code.infinito.nexus:2201/kevinveenbirkenbach/backup-docker-to-local.git
-https://pypi.org/project/baudolo/
+https://pypi.org/project/backup-docker-to-local/

Makefile

@@ -1,5 +1,6 @@
-.PHONY: install build \
-		test-e2e test test-unit test-integration
+.PHONY: install install-lint build clean lint ruff ruff-fix \
+		test test-unit test-integration test-e2e \
+		test-unit-run test-integration-run test-e2e-run
 
 # Default python if no venv is active
 PY_DEFAULT ?= python3
@@ -33,25 +34,49 @@ build:
 clean:
 	git clean -fdX .
 
-# ------------------------------------------------------------
-# Run E2E tests inside the container (Docker socket required)
-# ------------------------------------------------------------
-# E2E via isolated Docker-in-Docker (DinD)
-# - depends on local image build
-# - starts a DinD daemon container on a dedicated network
-# - loads the freshly built image into DinD
-# - runs the unittest suite inside a container that talks to DinD via DOCKER_HOST
-test-e2e: clean build
-	@bash scripts/test-e2e.sh
+# Separate from `install` so the test image does not have to carry the linter.
+install-lint:
+	@$(PY_DEFAULT) -m pip install -q -e ".[lint]"
 
-test: test-unit test-integration test-e2e
+# Runs on the host, not in the image, so it also covers what the Dockerfile
+# does not copy.
+ruff: install-lint
+	@echo ">> Running ruff over the whole repository"
+	@$(PY_DEFAULT) -m ruff check .
+	@$(PY_DEFAULT) -m ruff format --check .
 
-test-unit: clean build
+ruff-fix: install-lint
+	@$(PY_DEFAULT) -m ruff check --fix .
+	@$(PY_DEFAULT) -m ruff format .
+
+lint: ruff
+
+# build runs once, then lint and the three suites run concurrently via -j4; the
+# *-run targets carry no build prereq so the sub-make cannot race a second build.
+# `clean` is deliberately not a prerequisite; .dockerignore keeps the image
+# context clean instead.
+test:
+	@$(MAKE) build
+	@$(MAKE) -j4 lint test-unit-run test-integration-run test-e2e-run
+
+test-unit: build test-unit-run
+
+test-integration: build test-integration-run
+
+test-e2e: build test-e2e-run
+
+test-unit-run:
 	@echo ">> Running unit tests"
 	@docker run --rm -t $(IMAGE) \
-	  sh -lc 'python -m unittest discover -t . -s tests/unit -p "test_*.py" -v'
+	  bash -lc 'python -m unittest discover -t . -s tests/unit -p "test_*.py" -v'
 
-test-integration: clean build
+test-integration-run:
 	@echo ">> Running integration tests"
 	@docker run --rm -t $(IMAGE) \
-	  sh -lc 'python -m unittest discover -t . -s tests/integration -p "test_*.py" -v'
+	  bash -lc 'python -m unittest discover -t . -s tests/integration -p "test_*.py" -v'
+
+# E2E via isolated Docker-in-Docker (DinD): starts a DinD daemon on a dedicated
+# network, loads the freshly built image into it, and runs tests/e2e inside a
+# container that talks to DinD via DOCKER_HOST.
+test-e2e-run:
+	@bash scripts/test-e2e.sh

README.md

@@ -123,6 +123,8 @@ This information is used by `baudolo` to execute
 ```bash
 baudolo \
   --compose-dir /srv/docker \
+  --backups-dir /Backups \
+  --repo-name my-repo \
   --databases-csv /etc/baudolo/databases.csv \
   --database-containers central-postgres central-mariadb \
   --images-no-stop-required alpine postgres mariadb mysql \
@@ -133,11 +135,12 @@ baudolo \
 
 | Flag            | Description                                 |
 | --------------- | ------------------------------------------- |
-| `--everything`  | Always stop containers and re-run rsync     |
-| `--dump-only`   | Only create SQL dumps, skip file backups    |
+| `--only-sql`    | Skip file backups only for DB volumes when dumps succeed; non-DB volumes are still backed up; fallback to files if no dump.    |
+| `--only-files`  | Take no dumps at all; every volume is backed up as files. Needs no `--databases-csv`. Mutually exclusive with `--only-sql`. |
 | `--shutdown`    | Do not restart containers after backup      |
-| `--backups-dir` | Backup root directory (default: `/Backups`) |
-| `--repo-name`   | Backup namespace under machine hash         |
+| `--backups-dir` | Backup root directory (required)            |
+| `--repo-name`   | Backup namespace under machine hash (required) |
+| `--databases-csv`| Path to `databases.csv` (required)         |
 
 ## ♻️ Restore Operations
 

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "backup-docker-to-local"
-version = "1.0.0"
+version = "6.0.0"
 description = "Backup Docker volumes to local with rsync and optional DB dumps."
 readme = "README.md"
 requires-python = ">=3.9"
@@ -16,6 +16,11 @@ dependencies = [
   "dirval",
 ]
 
+[project.optional-dependencies]
+# Pinned: a ruff minor bump changes which rules fire, and `make test` gates on
+# a clean run, so an unpinned lint would fail the suite on an unrelated day.
+lint = ["ruff==0.16.1"]
+
 [project.scripts]
 baudolo = "baudolo.backup.__main__:main"
 baudolo-restore = "baudolo.restore.__main__:main"
@@ -27,3 +32,55 @@ package-dir = { "" = "src" }
 [tool.setuptools.packages.find]
 where = ["src"]
 exclude = ["tests*"]
+
+[tool.setuptools.package-data]
+"baudolo.restore.db" = ["*.sql"]
+
+[tool.ruff]
+respect-gitignore = true
+# The package still declares >=3.9, so pyupgrade must not propose 3.10+ syntax.
+target-version = "py39"
+exclude = ["build", "dist", "*.egg-info", ".venv", "venv"]
+
+[tool.ruff.lint]
+# Adopted from infinito-nexus-core so both repositories are held to one bar;
+# see that project's pyproject.toml for what each selector buys.
+select = [
+    "E", "F", "I", "B", "UP", "RUF", "SIM", "C4", "PERF", "RET", "PIE",
+    "T10", "PGH", "EXE", "RSE", "ICN", "DTZ",
+    "TID", "LOG", "G",
+    "S",
+    "PTH",
+    "FURB", "W", "FA", "YTT", "A", "ISC", "SLOT", "FLY",
+    "PYI",
+    "TC", "N",
+    "PLE0605",
+    "PLW1510", "PLW2901", "PLW0108", "PLW0603",
+    "PLR5501", "PLC0207", "PLR1722", "PLR1714",
+    "TRY002", "TRY004", "TRY300", "TRY301",
+    "BLE001",
+]
+
+# E501: `ruff format` reflows what it can; the rest is unsplittable literals.
+# RUF001/002/003: the prose uses em-dashes deliberately, not homoglyphs.
+# S603/S607: this tool's whole job is running `docker` / dump binaries from
+#   PATH in list form, which is already injection-safe.
+# PTH207/PTH208: changing `glob.glob`/`os.listdir` return shapes needs a
+#   per-call-site review, not a blanket rewrite.
+ignore = [
+    "E501",
+    "RUF001", "RUF002", "RUF003",
+    "S603", "S607",
+    "PTH207", "PTH208",
+]
+
+[tool.ruff.lint.per-file-ignores]
+# Test code legitimately uses what flake8-bandit flags in production code:
+# asserts, dummy credentials, /tmp fixtures, broad excepts in teardown, and
+# SQL built from fixture names (S608) to set the databases under test up.
+"tests/**" = [
+    "S101", "S102", "S105", "S106", "S108", "S110", "S112", "S608", "BLE001",
+]
+
+# The e2e helpers package is a deliberate re-export aggregator.
+"tests/e2e/helpers/__init__.py" = ["F403"]

scripts/test-e2e.sh

@@ -16,19 +16,30 @@ DIND="${E2E_DIND_NAME:-baudolo-e2e-dind}"
 DIND_VOL="${E2E_DIND_VOL:-baudolo-e2e-dind-data}"
 E2E_TMP_VOL="${E2E_TMP_VOL:-baudolo-e2e-tmp}"
 
-DIND_HOST="${E2E_DIND_HOST:-tcp://127.0.0.1:2375}"
+# Host-side access to the DinD daemon goes through `docker exec` (dind()
+# below) instead of a host-published port: port publishing is not reachable
+# from every environment (sandboxed runners, hosts with broken loopback
+# publishing), while exec only needs the outer docker socket. The TCP
+# listener stays for the test container inside the dedicated network.
 DIND_HOST_IN_NET="${E2E_DIND_HOST_IN_NET:-tcp://${DIND}:2375}"
 
+dind() { docker exec "${DIND}" docker "$@"; }
+dind_stdin() { docker exec -i "${DIND}" docker "$@"; }
+
 IMG="${E2E_IMAGE:-baudolo:local}"
-RSYNC_IMG="${E2E_RSYNC_IMAGE:-ghcr.io/kevinveenbirkenbach/alpine-rsync}"
 
 READY_TIMEOUT_SECONDS="${E2E_READY_TIMEOUT_SECONDS:-120}"
 ARTIFACTS_DIR="${E2E_ARTIFACTS_DIR:-./artifacts}"
 
+DIND_MTU="${E2E_DIND_MTU:-1280}"
+
 KEEP_ON_FAIL="${E2E_KEEP_ON_FAIL:-0}"
 KEEP_VOLUMES="${E2E_KEEP_VOLUMES:-0}"
 DEBUG_SHELL="${E2E_DEBUG_SHELL:-0}"
 
+# Override to run a subset, e.g. E2E_TEST_PATTERN=test_e2e_postgres_empty_drop_hard.py
+TEST_PATTERN="${E2E_TEST_PATTERN:-test_*.py}"
+
 FAILED=0
 TS="$(date +%Y%m%d%H%M%S)"
 
@@ -46,8 +57,8 @@ dump_debug() {
     echo "=== Host docker info ==="
     docker info || true
     echo
-    echo "=== DinD reachable? (docker -H ${DIND_HOST} version) ==="
-    docker -H "${DIND_HOST}" version || true
+    echo "=== DinD reachable? (docker exec ${DIND} docker version) ==="
+    dind version || true
     echo
   } > "${ARTIFACTS_DIR}/debug-host-${TS}.txt" 2>&1 || true
 
@@ -56,48 +67,34 @@ dump_debug() {
 
   # DinD state
   {
-    echo "=== docker -H ps -a ==="
-    docker -H "${DIND_HOST}" ps -a || true
+    echo "=== dind ps -a ==="
+    dind ps -a || true
     echo
-    echo "=== docker -H images ==="
-    docker -H "${DIND_HOST}" images || true
+    echo "=== dind images ==="
+    dind images || true
     echo
-    echo "=== docker -H network ls ==="
-    docker -H "${DIND_HOST}" network ls || true
+    echo "=== dind network ls ==="
+    dind network ls || true
     echo
-    echo "=== docker -H volume ls ==="
-    docker -H "${DIND_HOST}" volume ls || true
+    echo "=== dind volume ls ==="
+    dind volume ls || true
     echo
-    echo "=== docker -H system df ==="
-    docker -H "${DIND_HOST}" system df || true
+    echo "=== dind system df ==="
+    dind system df || true
   } > "${ARTIFACTS_DIR}/debug-dind-${TS}.txt" 2>&1 || true
 
   # Try to capture recent events (best effort; might be noisy)
-  docker -H "${DIND_HOST}" events --since 10m --until 0s \
+  dind events --since 10m --until 0s \
     > "${ARTIFACTS_DIR}/dind-events-${TS}.txt" 2>&1 || true
 
-  # Dump shared /tmp content from the tmp volume:
-  # We create a temporary container that mounts the volume, then tar its content.
-  # (Does not rely on host filesystem paths.)
+  # The shared tmp volume is mounted at /tmp inside the DinD container
+  # itself, so tar it there and copy it out with the outer daemon.
   log "DEBUG: archiving shared /tmp (volume ${E2E_TMP_VOL})"
-  docker -H "${DIND_HOST}" run --rm \
-    -v "${E2E_TMP_VOL}:/tmp" \
-    alpine:3.20 \
-    sh -lc 'cd /tmp && tar -czf /out.tar.gz . || true' \
-    >/dev/null 2>&1 || true
-
-  # The above writes inside the container FS, not to host. So do it properly:
-  # Use "docker cp" from a temp container.
-  local tmpc="baudolo-e2e-tmpdump-${TS}"
-  docker -H "${DIND_HOST}" rm -f "${tmpc}" >/dev/null 2>&1 || true
-  docker -H "${DIND_HOST}" create --name "${tmpc}" -v "${E2E_TMP_VOL}:/tmp" alpine:3.20 \
-    sh -lc 'cd /tmp && tar -czf /tmpdump.tar.gz . || true' >/dev/null
-  docker -H "${DIND_HOST}" start -a "${tmpc}" >/dev/null 2>&1 || true
-  docker -H "${DIND_HOST}" cp "${tmpc}:/tmpdump.tar.gz" "${ARTIFACTS_DIR}/e2e-tmp-${TS}.tar.gz" >/dev/null 2>&1 || true
-  docker -H "${DIND_HOST}" rm -f "${tmpc}" >/dev/null 2>&1 || true
+  docker exec "${DIND}" tar -czf "/tmpdump-${TS}.tar.gz" -C /tmp . >/dev/null 2>&1 || true
+  docker cp "${DIND}:/tmpdump-${TS}.tar.gz" "${ARTIFACTS_DIR}/e2e-tmp-${TS}.tar.gz" >/dev/null 2>&1 || true
 
   log "DEBUG: artifacts written:"
-  ls -la "${ARTIFACTS_DIR}" | sed 's/^/   /' || true
+  find "${ARTIFACTS_DIR}" -maxdepth 1 -mindepth 1 -print | sed 's/^/   /' || true
 }
 
 cleanup() {
@@ -105,9 +102,9 @@ cleanup() {
     log "KEEP_ON_FAIL=1 and failure detected -> skipping cleanup."
     log "Next steps:"
     echo "   - Inspect DinD logs: docker logs ${DIND} | less"
-    echo "   - Use DinD daemon:  docker -H ${DIND_HOST} ps -a"
-    echo "   - Shared tmp vol:   docker -H ${DIND_HOST} run --rm -v ${E2E_TMP_VOL}:/tmp alpine:3.20 ls -la /tmp"
-    echo "   - DinD docker root: docker -H ${DIND_HOST} run --rm -v ${DIND_VOL}:/var/lib/docker alpine:3.20 ls -la /var/lib/docker/volumes"
+    echo "   - Use DinD daemon:  docker exec ${DIND} docker ps -a"
+    echo "   - Shared tmp vol:   docker exec ${DIND} ls -la /tmp"
+    echo "   - DinD docker root: docker exec ${DIND} ls -la /var/lib/docker/volumes"
     return 0
   fi
 
@@ -124,8 +121,11 @@ cleanup() {
 }
 trap cleanup EXIT INT TERM
 
-log "Creating network ${NET} (if missing)"
-docker network inspect "${NET}" >/dev/null 2>&1 || docker network create "${NET}" >/dev/null
+log "(Re)creating network ${NET} with MTU ${DIND_MTU}"
+docker network rm "${NET}" >/dev/null 2>&1 || true
+docker network create \
+  --opt com.docker.network.driver.mtu="${DIND_MTU}" \
+  "${NET}" >/dev/null
 
 log "Removing old ${DIND} (if any)"
 docker rm -f "${DIND}" >/dev/null 2>&1 || true
@@ -145,14 +145,14 @@ docker run -d --privileged \
   -e DOCKER_TLS_CERTDIR="" \
   -v "${DIND_VOL}:/var/lib/docker" \
   -v "${E2E_TMP_VOL}:/tmp" \
-  -p 2375:2375 \
   docker:dind \
   --host=tcp://0.0.0.0:2375 \
-  --tls=false >/dev/null
+  --tls=false \
+  --mtu="${DIND_MTU}" >/dev/null
 
 log "Waiting for DinD to be ready..."
 for i in $(seq 1 "${READY_TIMEOUT_SECONDS}"); do
-  if docker -H "${DIND_HOST}" version >/dev/null 2>&1; then
+  if dind version >/dev/null 2>&1; then
     log "DinD is ready."
     break
   fi
@@ -166,15 +166,11 @@ for i in $(seq 1 "${READY_TIMEOUT_SECONDS}"); do
   fi
 done
 
-log "Pre-pulling helper images in DinD..."
-log " - Pulling: ${RSYNC_IMG}"
-docker -H "${DIND_HOST}" pull "${RSYNC_IMG}"
-
 log "Ensuring alpine exists in DinD (for debug helpers)"
-docker -H "${DIND_HOST}" pull alpine:3.20 >/dev/null
+dind pull alpine:3.20 >/dev/null
 
 log "Loading ${IMG} image into DinD..."
-docker save "${IMG}" | docker -H "${DIND_HOST}" load >/dev/null
+docker save "${IMG}" | dind_stdin load >/dev/null
 
 log "Running E2E tests inside DinD"
 set +e
@@ -183,11 +179,10 @@ if [ "${DEBUG_SHELL}" = "1" ]; then
   docker run --rm -it \
     --network "${NET}" \
     -e DOCKER_HOST="${DIND_HOST_IN_NET}" \
-    -e E2E_RSYNC_IMAGE="${RSYNC_IMG}" \
-    -v "${DIND_VOL}:/var/lib/docker:ro" \
+    -v "${DIND_VOL}:/var/lib/docker" \
     -v "${E2E_TMP_VOL}:/tmp" \
     "${IMG}" \
-    sh -lc '
+    bash -lc '
       set -e
       if [ ! -f /etc/machine-id ]; then
         mkdir -p /etc
@@ -195,18 +190,19 @@ if [ "${DEBUG_SHELL}" = "1" ]; then
       fi
       echo ">> DOCKER_HOST=${DOCKER_HOST}"
       docker ps -a || true
-      exec sh
+      exec bash
     '
   rc=$?
 else
   docker run --rm \
     --network "${NET}" \
     -e DOCKER_HOST="${DIND_HOST_IN_NET}" \
-    -e E2E_RSYNC_IMAGE="${RSYNC_IMG}" \
-    -v "${DIND_VOL}:/var/lib/docker:ro" \
+    -e E2E_TEST_PATTERN="${TEST_PATTERN}" \
+    -e E2E_REQUIRE_FILESYSTEMS="${E2E_REQUIRE_FILESYSTEMS:-}" \
+    -v "${DIND_VOL}:/var/lib/docker" \
     -v "${E2E_TMP_VOL}:/tmp" \
     "${IMG}" \
-    sh -lc '
+    bash -lc '
     set -euo pipefail
     set -x
     export PYTHONUNBUFFERED=1
@@ -218,7 +214,7 @@ else
         cat /proc/sys/kernel/random/uuid > /etc/machine-id
     fi
 
-    python -m unittest discover -t . -s tests/e2e -p "test_*.py" -v -f
+    python -m unittest discover -t . -s tests/e2e -p "${E2E_TEST_PATTERN}" -v -f
     '
   rc=$?
 fi

src/baudolo/backup/__main__.py

@@ -1,9 +1,6 @@
-#!/usr/bin/env python3
-
 from __future__ import annotations
 
 from .app import main
 
-
 if __name__ == "__main__":
     raise SystemExit(main())

src/baudolo/backup/app.py

@@ -1,183 +1,143 @@
+"""Back up every Docker volume of a host into a timestamped generation."""
+
 from __future__ import annotations
 
-import os
-import pathlib
+from contextlib import ExitStack
 from datetime import datetime
-
-import pandas
-from dirval import create_stamp_file
+from pathlib import Path
 
 from .cli import parse_args
 from .compose import handle_docker_compose_services
-from .db import backup_database
 from .docker import (
     change_containers_status,
     containers_using_volume,
     docker_volume_names,
-    get_image_info,
-    has_image,
+    filter_stoppable,
 )
-from .shell import execute_shell_command
-from .volume import backup_volume
-
-
-def get_machine_id() -> str:
-    return execute_shell_command("sha256sum /etc/machine-id")[0][0:64]
-
-
-def stamp_directory(version_dir: str) -> None:
-    """
-    Use dirval as a Python library to stamp the directory (no CLI dependency).
-    """
-    create_stamp_file(version_dir)
-
-
-def create_version_directory(versions_dir: str, backup_time: str) -> str:
-    version_dir = os.path.join(versions_dir, backup_time)
-    pathlib.Path(version_dir).mkdir(parents=True, exist_ok=True)
-    return version_dir
-
-
-def create_volume_directory(version_dir: str, volume_name: str) -> str:
-    path = os.path.join(version_dir, volume_name)
-    pathlib.Path(path).mkdir(parents=True, exist_ok=True)
-    return path
-
-
-def is_image_ignored(container: str, images_no_backup_required: list[str]) -> bool:
-    if not images_no_backup_required:
-        return False
-    img = get_image_info(container)
-    return any(pat in img for pat in images_no_backup_required)
-
-
-def volume_is_fully_ignored(containers: list[str], images_no_backup_required: list[str]) -> bool:
-    """
-    Skip file backup only if all containers linked to the volume are ignored.
-    """
-    if not containers:
-        return False
-    return all(is_image_ignored(c, images_no_backup_required) for c in containers)
-
-
-def requires_stop(containers: list[str], images_no_stop_required: list[str]) -> bool:
-    """
-    Stop is required if ANY container image is NOT in the whitelist patterns.
-    """
-    for c in containers:
-        img = get_image_info(c)
-        if not any(pat in img for pat in images_no_stop_required):
-            return True
-    return False
-
-
-def backup_mariadb_or_postgres(
-    *,
-    container: str,
-    volume_dir: str,
-    databases_df: "pandas.DataFrame",
-    database_containers: list[str],
-) -> bool:
-    """
-    Returns True if the container is a DB container we handled.
-    """
-    for img in ["mariadb", "postgres"]:
-        if has_image(container, img):
-            backup_database(
-                container=container,
-                volume_dir=volume_dir,
-                db_type=img,
-                databases_df=databases_df,
-                database_containers=database_containers,
-            )
-            return True
-    return False
-
-
-def _backup_dumps_for_volume(
-    *,
-    containers: list[str],
-    vol_dir: str,
-    databases_df: "pandas.DataFrame",
-    database_containers: list[str],
-) -> bool:
-    """
-    Create DB dumps for any mariadb/postgres containers attached to this volume.
-    Returns True if at least one dump was produced.
-    """
-    dumped_any = False
-    for c in containers:
-        if backup_mariadb_or_postgres(
-            container=c,
-            volume_dir=vol_dir,
-            databases_df=databases_df,
-            database_containers=database_containers,
-        ):
-            dumped_any = True
-    return dumped_any
+from .dumps import VolumeOutcome, backup_dumps_for_volume, load_databases_df
+from .layout import (
+    create_version_directory,
+    create_volume_directory,
+    get_machine_id,
+    stamp_directory,
+    write_manifest,
+)
+from .policy import requires_stop, volume_is_fully_ignored
+from .snapshot import snapshot_source, volume_snapshot
+from .volume import backup_volume, inspect_backing
 
 
 def main() -> int:
     args = parse_args()
 
     machine_id = get_machine_id()
-    backup_time = datetime.now().strftime("%Y%m%d%H%M%S")
+    # Local wall clock on purpose: generations sort by this name, and UTC would
+    # order new ones before the existing ones wherever the offset is positive.
+    backup_time = datetime.now().strftime("%Y%m%d%H%M%S")  # noqa: DTZ005
 
-    versions_dir = os.path.join(args.backups_dir, machine_id, args.repo_name)
+    versions_dir = str(Path(args.backups_dir) / machine_id / args.repo_name)
     version_dir = create_version_directory(versions_dir, backup_time)
 
-    databases_df = pandas.read_csv(args.databases_csv, sep=";")
+    databases_df = None if args.only_files else load_databases_df(args.databases_csv)
 
     print("💾 Start volume backups...", flush=True)
 
-    for volume_name in docker_volume_names():
-        print(f"Start backup routine for volume: {volume_name}", flush=True)
-        containers = containers_using_volume(volume_name)
+    outcomes: dict[str, VolumeOutcome] = {}
 
-        vol_dir = create_volume_directory(version_dir, volume_name)
-
-        # Old behavior: DB dumps are additional to file backups.
-        _backup_dumps_for_volume(
-            containers=containers,
-            vol_dir=vol_dir,
-            databases_df=databases_df,
-            database_containers=args.database_containers,
-        )
-
-        # dump-only: skip ALL file rsync backups
-        if args.dump_only:
-            continue
-
-        # skip file backup if all linked containers are ignored
-        if volume_is_fully_ignored(containers, args.images_no_backup_required):
-            print(
-                f"Skipping file backup for volume '{volume_name}' (all linked containers are ignored).",
-                flush=True,
+    with ExitStack() as stack:
+        resolve_source = None
+        if args.snapshot:
+            resolve_source = stack.enter_context(
+                volume_snapshot(args.snapshot, args.snapshot_subject, backup_time)
             )
-            continue
 
-        if args.everything:
-            # "everything": always do pre-rsync, then stop + rsync again
-            backup_volume(versions_dir, volume_name, vol_dir)
-            change_containers_status(containers, "stop")
-            backup_volume(versions_dir, volume_name, vol_dir)
-            if not args.shutdown:
-                change_containers_status(containers, "start")
-            continue
+        for volume_name in docker_volume_names():
+            print(f"Start backup routine for volume: {volume_name}", flush=True)
 
-        # default: rsync, and if needed stop + rsync
-        backup_volume(versions_dir, volume_name, vol_dir)
-        if requires_stop(containers, args.images_no_stop_required):
-            change_containers_status(containers, "stop")
-            backup_volume(versions_dir, volume_name, vol_dir)
-            if not args.shutdown:
-                change_containers_status(containers, "start")
+            if volume_name in args.volumes_no_backup_required:
+                print(
+                    f"Skipping volume '{volume_name}' entirely (declared no-backup).",
+                    flush=True,
+                )
+                continue
 
-    # Stamp the backup version directory using dirval (python lib)
+            containers = containers_using_volume(volume_name)
+
+            if volume_is_fully_ignored(containers, args.images_no_backup_required):
+                print(
+                    f"Skipping volume '{volume_name}' entirely (all linked containers are ignored).",
+                    flush=True,
+                )
+                continue
+
+            vol_dir = create_volume_directory(version_dir, volume_name)
+
+            outcome = VolumeOutcome(database=False, dumped=False)
+            if not args.only_files:
+                outcome = backup_dumps_for_volume(
+                    containers=containers,
+                    vol_dir=vol_dir,
+                    databases_df=databases_df,
+                    database_containers=args.database_containers,
+                )
+            outcomes[volume_name] = outcome
+
+            if args.only_sql and outcome.database:
+                if not outcome.dumped:
+                    print(
+                        f"WARNING: only-sql requested but no DB dump was produced for DB volume '{volume_name}'. "
+                        "Falling back to file backup.",
+                        flush=True,
+                    )
+                else:
+                    continue
+
+            backing = inspect_backing(volume_name)
+            live_source = backing.source
+
+            def copy(
+                *,
+                authoritative: bool,
+                source: str = live_source,
+                volume: str = volume_name,
+                target: str = vol_dir,
+            ) -> None:
+                backup_volume(
+                    versions_dir,
+                    volume,
+                    target,
+                    authoritative=authoritative,
+                    source=source,
+                )
+
+            if resolve_source is not None:
+                source, reason = snapshot_source(
+                    resolve_source, backing, args.snapshot_subject
+                )
+                if source is not None:
+                    copy(authoritative=True, source=source)
+                else:
+                    print(
+                        f"WARNING: volume '{volume_name}' is not in the snapshot "
+                        f"({reason}); copying it live instead.",
+                        flush=True,
+                    )
+                    copy(authoritative=False)
+                continue
+
+            copy(authoritative=False)
+            if requires_stop(containers, args.images_no_stop_required):
+                stoppable = filter_stoppable(containers)
+                change_containers_status(stoppable, "stop")
+                copy(authoritative=True)
+                if not args.shutdown:
+                    change_containers_status(stoppable, "start")
+
+    write_manifest(version_dir, outcomes)
     stamp_directory(version_dir)
     print("Finished volume backups.", flush=True)
 
     print("Handling Docker Compose services...", flush=True)
-    handle_docker_compose_services(args.compose_dir, args.docker_compose_hard_restart_required)
+    handle_docker_compose_services(args.compose_dir, args.hard_restart_projects)
 
     return 0

src/baudolo/backup/cli.py

@@ -1,29 +1,9 @@
 from __future__ import annotations
 
 import argparse
-import os
-from pathlib import Path
-
-
-def _default_repo_name() -> str:
-    """
-    Derive the repository name from the folder that contains `src/`.
-
-    Expected layout:
-      <repo-root>/src/baudolo/backup/cli.py
-
-    => parents[0]=backup, [1]=baudolo, [2]=src, [3]=repo-root
-    """
-    try:
-        return Path(__file__).resolve().parents[3].name
-    except Exception:
-        return "backup-docker-to-local"
 
 
 def parse_args() -> argparse.Namespace:
-    dirname = os.path.dirname(__file__)
-    default_databases_csv = os.path.join(dirname, "databases.csv")
-
     p = argparse.ArgumentParser(description="Backup Docker volumes.")
 
     p.add_argument(
@@ -33,61 +13,102 @@ def parse_args() -> argparse.Namespace:
         help="Path to the parent directory containing docker-compose setups",
     )
     p.add_argument(
-        "--docker-compose-hard-restart-required",
-        nargs="+",
-        default=["mailu"],
-        help="Compose dir names that require 'docker-compose down && up -d' (default: mailu)",
+        "--hard-restart-projects",
+        nargs="*",
+        default=[],
+        help="Compose dir names that require 'docker-compose down && up -d' (default: none; pass e.g. 'mailu' under compose where the DB cannot be backed up hot)",
     )
 
     p.add_argument(
         "--repo-name",
-        default=_default_repo_name(),
-        help="Backup repo folder name under <backups-dir>/<machine-id>/ (default: git repo folder name)",
+        required=True,
+        help="Backup repo folder name under <backups-dir>/<machine-id>/",
     )
     p.add_argument(
         "--databases-csv",
-        default=default_databases_csv,
-        help=f"Path to databases.csv (default: {default_databases_csv})",
+        help="Path to databases.csv; required unless --only-files is given",
     )
     p.add_argument(
         "--backups-dir",
-        default="/Backups",
-        help="Backup root directory (default: /Backups)",
+        required=True,
+        help="Backup root directory (e.g. /var/lib/backup/)",
+    )
+
+    p.add_argument(
+        "--snapshot",
+        choices=["btrfs", "zfs"],
+        help="Capture every volume from one atomic filesystem snapshot instead of copying the live tree. Containers are not stopped, and the copy is a single pass. Requires --snapshot-subject. Omit to keep the live two-pass copy.",
+    )
+    p.add_argument(
+        "--snapshot-subject",
+        help="Btrfs subvolume or zfs dataset mountpoint holding the docker volumes, e.g. /var/lib/docker. Required with --snapshot.",
     )
 
     p.add_argument(
         "--database-containers",
         nargs="+",
-        required=True,
+        default=[],
         help="Container names treated as special instances for database backups",
     )
     p.add_argument(
         "--images-no-stop-required",
         nargs="+",
-        required=True,
-        help="Image name patterns for which containers should not be stopped during file backup",
+        default=[],
+        help="Exact image references (repo:tag, incl. any registry prefix) whose containers must not be stopped during file backup",
     )
     p.add_argument(
         "--images-no-backup-required",
         nargs="+",
         default=[],
-        help="Image name patterns for which no backup should be performed",
+        help="Exact image references (repo:tag, incl. any registry prefix) for which no backup should be performed",
     )
 
     p.add_argument(
-        "--everything",
-        action="store_true",
-        help="Force file backup for all volumes and also execute database dumps (like old script)",
+        "--volumes-no-backup-required",
+        nargs="+",
+        default=[],
+        help="Exact volume names that are never backed up, whatever containers use them. For derived trees a restore cannot reproduce, above all a nested docker data root",
     )
+
     p.add_argument(
         "--shutdown",
         action="store_true",
         help="Do not restart containers after backup",
     )
-    p.add_argument(
-        "--dump-only",
-        action="store_true",
-        help="Only create DB dumps (skip ALL file rsync backups)",
-    )
 
-    return p.parse_args()
+    scope = p.add_mutually_exclusive_group()
+    scope.add_argument(
+        "--only-sql",
+        action="store_true",
+        help=(
+            "Create database dumps only for DB volumes. "
+            "File backups are skipped for DB volumes if a dump succeeds, "
+            "but non-DB volumes are still backed up. "
+            "If a DB dump cannot be produced, baudolo falls back to a file backup."
+        ),
+    )
+    scope.add_argument(
+        "--only-files",
+        action="store_true",
+        help=(
+            "Take no database dumps at all and back up every volume as files. "
+            "For hosts that hold no database credentials. A database's files "
+            "are only consistent if its containers are stopped for the second "
+            "pass, so keep its image off --images-no-stop-required."
+        ),
+    )
+    args = p.parse_args()
+    if not args.only_files and not args.databases_csv:
+        p.error("--databases-csv is required unless --only-files is given")
+    if bool(args.snapshot) != bool(args.snapshot_subject):
+        p.error("--snapshot and --snapshot-subject must be given together")
+    if args.snapshot and args.shutdown:
+        p.error(
+            "--shutdown is meaningless with --snapshot: containers are never stopped"
+        )
+    if args.snapshot and args.hard_restart_projects:
+        p.error(
+            "--hard-restart-projects is meaningless with --snapshot: the flag exists "
+            "for stacks whose database cannot be backed up hot, which a snapshot solves"
+        )
+    return args

src/baudolo/backup/compose.py

@@ -1,27 +1,94 @@
 from __future__ import annotations
 
 import os
+import shutil
 import subprocess
+from pathlib import Path
+
+
+def _build_compose_cmd(project_dir: str, passthrough: list[str]) -> list[str]:
+    """
+    Build the compose command for this project directory.
+
+    Policy:
+    - If `compose` wrapper exists (Infinito.Nexus): use it and delegate ALL logic to it.
+    - Else: use plain `docker compose` with --chdir.
+    - NO custom compose file/env detection in this project.
+    """
+    pdir = Path(project_dir).resolve()
+
+    wrapper = shutil.which("compose")
+    if wrapper:
+        # "--" ensures wrapper stops parsing its own args.
+        return [wrapper, "--chdir", str(pdir), "--", *passthrough]
+
+    docker = shutil.which("docker")
+    if docker:
+        return [docker, "compose", "--chdir", str(pdir), *passthrough]
+
+    raise RuntimeError("Neither 'compose' nor 'docker' found in PATH")
+
+
+def _find_compose_file(project_dir: str) -> Path | None:
+    """
+    Detect a compose file in `project_dir` (case-insensitive).
+
+    Supported names:
+    - compose.yml / compose.yaml
+    - docker-compose.yml / docker-compose.yaml
+    """
+    pdir = Path(project_dir)
+    if not pdir.is_dir():
+        return None
+
+    # Map lowercase filename -> actual Path (preserves original casing)
+    by_lower = {p.name.lower(): p for p in pdir.iterdir() if p.is_file()}
+
+    # Preferred order (policy decision)
+    candidates = [
+        "docker-compose.yml",
+        "docker-compose.yaml",
+        "compose.yml",
+        "compose.yaml",
+    ]
+
+    for name in candidates:
+        found = by_lower.get(name)
+        if found is not None:
+            return found
+
+    return None
 
 
 def hard_restart_docker_services(dir_path: str) -> None:
-    print(f"Hard restart docker-compose services in: {dir_path}", flush=True)
-    subprocess.run(["docker-compose", "down"], cwd=dir_path, check=True)
-    subprocess.run(["docker-compose", "up", "-d"], cwd=dir_path, check=True)
+    print(f"Hard restart compose services in: {dir_path}", flush=True)
+
+    down_cmd = _build_compose_cmd(dir_path, ["down"])
+    up_cmd = _build_compose_cmd(dir_path, ["up", "-d"])
+
+    print(">>> " + " ".join(down_cmd), flush=True)
+    subprocess.run(down_cmd, check=True)
+
+    print(">>> " + " ".join(up_cmd), flush=True)
+    subprocess.run(up_cmd, check=True)
 
 
-def handle_docker_compose_services(parent_directory: str, hard_restart_required: list[str]) -> None:
+def handle_docker_compose_services(
+    parent_directory: str,
+    hard_restart_required: list[str],
+) -> None:
     for entry in os.scandir(parent_directory):
         if not entry.is_dir():
             continue
 
         dir_path = entry.path
-        name = os.path.basename(dir_path)
-        compose_file = os.path.join(dir_path, "docker-compose.yml")
+        name = Path(dir_path).name
 
         print(f"Checking directory: {dir_path}", flush=True)
-        if not os.path.isfile(compose_file):
-            print("No docker-compose.yml found. Skipping.", flush=True)
+
+        compose_file = _find_compose_file(dir_path)
+        if compose_file is None:
+            print("No supported compose file found. Skipping.", flush=True)
             continue
 
         if name in hard_restart_required:

src/baudolo/backup/db.py

@@ -1,26 +1,61 @@
 from __future__ import annotations
 
-import os
+import logging
 import pathlib
 import re
+from typing import TYPE_CHECKING
 
-import pandas
+from baudolo.databases import CLUSTER_ROW, validate_database
+from baudolo.generation import CLUSTER_SUFFIX, DUMP_SUFFIX, SQL_DIR
 
-from .shell import BackupException, execute_shell_command
+from .docker import docker_exec_argv
+from .shell import BackupError, execute_to_file
+
+if TYPE_CHECKING:
+    import pandas as pd
+
+log = logging.getLogger(__name__)
 
 
-def get_instance(container: str, database_containers: list[str]) -> str:
+def get_instance(container: str, database_containers: list[str]) -> str | None:
+    """The databases.csv instance a container serves, or None for no database.
+
+    A declared container is its own instance. Every other name is normalised by
+    stripping a database suffix token, which maps both `<app>-database` from
+    compose and `<app>_database.1.<task>` from swarm onto the same instance.
+
+    Args:
+        container: the running container's name.
+        database_containers: names passed via --database-containers, taken as
+            declared engines whatever they are called.
+
+    Returns:
+        The instance name, or None when the name carries no database token: an
+        application container is not an engine, even when it ships the client
+        tools that would let a dump command start.
+    """
     if container in database_containers:
         return container
-    return re.split(r"(_|-)(database|db|postgres)", container)[0]
+    parts = re.split(r"(_|-)(database|db|postgres)", container)
+    return parts[0] if len(parts) > 1 else None
 
 
-def fallback_pg_dumpall(container: str, username: str, password: str, out_file: str) -> None:
-    cmd = (
-        f"PGPASSWORD={password} docker exec -i {container} "
-        f"pg_dumpall -U {username} -h localhost > {out_file}"
+def fallback_pg_dumpall(
+    container: str, username: str, password: str, out_file: str
+) -> None:
+    """
+    Perform a full Postgres cluster dump using pg_dumpall.
+    """
+    execute_to_file(
+        docker_exec_argv(
+            container,
+            ["pg_dumpall", "-U", username, "-h", "localhost"],
+            interactive=True,
+            forward_env=["PGPASSWORD"],
+        ),
+        out_file,
+        env={"PGPASSWORD": password},
     )
-    execute_shell_command(cmd)
 
 
 def backup_database(
@@ -28,46 +63,107 @@ def backup_database(
     container: str,
     volume_dir: str,
     db_type: str,
-    databases_df: "pandas.DataFrame",
+    dump_tool: str,
+    databases_df: pd.DataFrame,
     database_containers: list[str],
-) -> None:
+) -> bool:
+    """
+    Backup databases for a given DB container.
+
+    Args:
+        dump_tool: the MariaDB client found in the container, so an image
+            that ships only mysqldump is dumped with the tool it has.
+
+    Returns True if at least one dump was produced.
+    """
     instance_name = get_instance(container, database_containers)
-    entries = databases_df.loc[databases_df["instance"] == instance_name]
+    if instance_name is None:
+        log.debug("Container '%s' carries no database token", container)
+        return False
+
+    entries = databases_df[databases_df["instance"] == instance_name]
     if entries.empty:
-        raise BackupException(f"No entry found for instance '{instance_name}'")
+        log.debug("No database entries for instance '%s'", instance_name)
+        return False
 
-    out_dir = os.path.join(volume_dir, "sql")
-    pathlib.Path(out_dir).mkdir(parents=True, exist_ok=True)
+    out_dir = pathlib.Path(volume_dir) / SQL_DIR
+    out_dir.mkdir(parents=True, exist_ok=True)
 
-    for row in entries.iloc:
-        db_name = row["database"]
-        user = row["username"]
-        password = row["password"]
+    produced = False
 
-        dump_file = os.path.join(out_dir, f"{db_name}.backup.sql")
+    for row in entries.itertuples(index=False):
+        raw_db = getattr(row, "database", "")
+        user = (getattr(row, "username", "") or "").strip()
+        password = (getattr(row, "password", "") or "").strip()
+
+        db_value = validate_database(raw_db, instance=instance_name)
+
+        if db_value == CLUSTER_ROW:
+            if db_type != "postgres":
+                raise ValueError(
+                    f"databases.csv entry for instance '{instance_name}': "
+                    f"'{CLUSTER_ROW}' is currently only supported for Postgres."
+                )
+
+            cluster_file = str(out_dir / f"{instance_name}{CLUSTER_SUFFIX}")
+            fallback_pg_dumpall(container, user, password, cluster_file)
+            produced = True
+            continue
+
+        db_name = db_value
+        dump_file = str(out_dir / f"{db_name}{DUMP_SUFFIX}")
 
         if db_type == "mariadb":
-            cmd = (
-                f"docker exec {container} /usr/bin/mariadb-dump "
-                f"-u {user} -p{password} {db_name} > {dump_file}"
+            # Force TCP so auth matches '<user>'@'%' instead of socket -> 'localhost'.
+            execute_to_file(
+                docker_exec_argv(
+                    container,
+                    [
+                        dump_tool,
+                        "-h",
+                        "127.0.0.1",
+                        "--protocol=tcp",
+                        "-u",
+                        user,
+                        f"-p{password}",
+                        db_name,
+                    ],
+                ),
+                dump_file,
             )
-            execute_shell_command(cmd)
+            produced = True
             continue
 
         if db_type == "postgres":
-            cluster_file = os.path.join(out_dir, f"{instance_name}.cluster.backup.sql")
-            if not db_name:
-                fallback_pg_dumpall(container, user, password, cluster_file)
-                return
-
             try:
-                cmd = (
-                    f"PGPASSWORD={password} docker exec -i {container} "
-                    f"pg_dump -U {user} -d {db_name} -h localhost > {dump_file}"
+                execute_to_file(
+                    docker_exec_argv(
+                        container,
+                        [
+                            "pg_dump",
+                            "-U",
+                            user,
+                            "-d",
+                            db_name,
+                            "-h",
+                            "localhost",
+                            "--no-owner",
+                            "--no-privileges",
+                        ],
+                        interactive=True,
+                        forward_env=["PGPASSWORD"],
+                    ),
+                    dump_file,
+                    env={"PGPASSWORD": password},
                 )
-                execute_shell_command(cmd)
-            except BackupException as e:
-                print(f"pg_dump failed: {e}", flush=True)
-                print(f"Falling back to pg_dumpall for instance '{instance_name}'", flush=True)
-                fallback_pg_dumpall(container, user, password, cluster_file)
+                produced = True
+            except BackupError as e:
+                raise BackupError(
+                    f"Postgres dump failed for instance '{instance_name}', "
+                    f"database '{db_name}'. This database was explicitly configured "
+                    "and therefore must succeed.\n"
+                    f"{e}"
+                ) from e
             continue
+
+    return produced

src/baudolo/backup/docker.py

@@ -1,43 +1,143 @@
 from __future__ import annotations
 
-from .shell import execute_shell_command
+from typing import TYPE_CHECKING
+
+from .shell import BackupError, execute_shell_command
+
+if TYPE_CHECKING:
+    from collections.abc import Sequence
+
+
+def docker_exec_argv(
+    container: str,
+    argv: Sequence[str],
+    *,
+    interactive: bool = False,
+    forward_env: Sequence[str] = (),
+) -> list[str]:
+    """The argv that runs *argv* inside *container*.
+
+    Args:
+        container: the container to run in.
+        argv: the command, already split.
+        interactive: keep stdin open, for a command that is fed a dump.
+        forward_env: names of environment variables to hand to the container.
+            Passed as bare ``-e NAME``, so docker copies the value out of this
+            process's own environment; spelling ``-e NAME=value`` instead would
+            publish a secret in the host's process list.
+
+    Returns:
+        The argv list.
+    """
+    forwarded = [arg for name in forward_env for arg in ("-e", name)]
+    return [
+        "docker",
+        "exec",
+        *(["-i"] if interactive else []),
+        *forwarded,
+        container,
+        *argv,
+    ]
 
 
 def get_image_info(container: str) -> str:
     return execute_shell_command(
-        f"docker inspect --format '{{{{.Config.Image}}}}' {container}"
+        ["docker", "inspect", "--format", "{{.Config.Image}}", container]
     )[0]
 
 
-def has_image(container: str, pattern: str) -> bool:
-    """Return True if container's image contains the pattern."""
-    return pattern in get_image_info(container)
+def image_id(container: str) -> str:
+    """The container's image ID, identical for every replica of one image."""
+    return execute_shell_command(
+        ["docker", "inspect", "--format", "{{.Image}}", container]
+    )[0].strip()
+
+
+def has_tool(container: str, tool: str) -> bool:
+    """Whether *tool* runs inside the container.
+
+    Executes the binary rather than asking a shell for it: a distroless image
+    has no shell, and `sh -c 'command -v'` would answer "absent" for every
+    tool it ships.
+    """
+    try:
+        execute_shell_command(docker_exec_argv(container, [tool, "--version"]))
+    except BackupError:
+        return False
+    return True
 
 
 def docker_volume_names() -> list[str]:
-    return execute_shell_command("docker volume ls --format '{{.Name}}'")
+    return execute_shell_command(["docker", "volume", "ls", "--format", "{{.Name}}"])
 
 
 def containers_using_volume(volume_name: str) -> list[str]:
     return execute_shell_command(
-        f"docker ps --filter volume=\"{volume_name}\" --format '{{{{.Names}}}}'"
+        [
+            "docker",
+            "ps",
+            "--filter",
+            f"volume={volume_name}",
+            "--format",
+            "{{.Names}}",
+        ]
     )
 
 
+def is_swarm_task(container: str) -> bool:
+    """Swarm-managed task containers must never be stopped or started
+    manually: the orchestrator replaces the stopped task and a later
+    `docker start` fails on the detached overlay network. A container that
+    vanished between listing and inspect (--rm one-shots, task-history GC)
+    counts as not stoppable instead of aborting the whole backup run; if the
+    container still exists the inspect failure re-raises, so a broken daemon
+    keeps failing the run loudly instead of silently skipping the stop."""
+    try:
+        out = execute_shell_command(
+            [
+                "docker",
+                "inspect",
+                "--format",
+                '{{index .Config.Labels "com.docker.swarm.task.id"}}',
+                container,
+            ]
+        )
+    except BackupError:
+        still_listed = execute_shell_command(
+            [
+                "docker",
+                "ps",
+                "-a",
+                "--filter",
+                f"name=^{container}$",
+                "--format",
+                "{{.Names}}",
+            ]
+        )
+        if still_listed and still_listed[0].strip():
+            raise
+        return True
+    return bool(out and out[0].strip())
+
+
+def filter_stoppable(containers: list[str]) -> list[str]:
+    """Containers baudolo may stop/start itself (everything but swarm tasks)."""
+    stoppable = []
+    for container in containers:
+        if is_swarm_task(container):
+            print(
+                f"Skipping stop/start for swarm task container '{container}'.",
+                flush=True,
+            )
+            continue
+        stoppable.append(container)
+    return stoppable
+
+
 def change_containers_status(containers: list[str], status: str) -> None:
     """Stop or start a list of containers."""
     if not containers:
         print(f"No containers to {status}.", flush=True)
         return
-    names = " ".join(containers)
-    print(f"{status.capitalize()} containers: {names}...", flush=True)
-    execute_shell_command(f"docker {status} {names}")
-
-
-def docker_volume_exists(volume: str) -> bool:
-    # Avoid throwing exceptions for exists checks.
-    try:
-        execute_shell_command(f"docker volume inspect {volume} >/dev/null 2>&1 && echo OK")
-        return True
-    except Exception:
-        return False
+    print(f"{status.capitalize()} containers: {' '.join(containers)}...", flush=True)
+    execute_shell_command(["docker", status, *containers])

src/baudolo/backup/dumps.py

@@ -0,0 +1,153 @@
+"""Database dumps taken before a volume's files are copied."""
+
+from __future__ import annotations
+
+import sys
+from typing import NamedTuple
+
+import pandas as pd
+from pandas.errors import EmptyDataError
+
+from baudolo.databases import COLUMNS, DELIMITER
+
+from .db import backup_database
+from .docker import has_tool, image_id
+
+DUMP_TOOLS: tuple[tuple[str, str], ...] = (
+    ("postgres", "pg_dumpall"),
+    ("mariadb", "mariadb-dump"),
+    ("mariadb", "mysqldump"),
+)
+
+_ENGINE_BY_IMAGE: dict[str, tuple[str, str] | None] = {}
+
+
+class VolumeOutcome(NamedTuple):
+    """What a dump attempt established about one volume.
+
+    ``database`` says a container serving the volume speaks an engine this
+    tool can dump; ``dumped`` says a dump was actually written. ``engine`` is
+    the engine that was detected, or None when none was.
+    """
+
+    database: bool
+    dumped: bool
+    engine: str | None = None
+
+
+def container_engine(container: str) -> tuple[str, str] | None:
+    """The (engine, dump tool) a container can serve, or None for neither.
+
+    Asks the container what it can run instead of reading its image name. A
+    dedicated Postgres is tagged `<app>-database` or `postgis/postgis` and
+    carries no engine token at all, while a swarm registry host such as
+    `svc-db-mariadb-swarm-mgr-01:5000` carries the wrong one.
+
+    Args:
+        container: must be running - `docker exec` is the probe, and a
+            stopped container would be cached as "no engine" for its whole
+            image. The only caller feeds it `docker ps` output.
+
+    Returns:
+        The engine and the tool that dumps it, cached per image ID so that
+        replicas of one image are probed once.
+    """
+    image = image_id(container)
+    if image not in _ENGINE_BY_IMAGE:
+        _ENGINE_BY_IMAGE[image] = next(
+            (
+                (engine, tool)
+                for engine, tool in DUMP_TOOLS
+                if has_tool(container, tool)
+            ),
+            None,
+        )
+    return _ENGINE_BY_IMAGE[image]
+
+
+def backup_mariadb_or_postgres(
+    *,
+    container: str,
+    volume_dir: str,
+    databases_df: pd.DataFrame,
+    database_containers: list[str],
+) -> VolumeOutcome:
+    """What this container contributes to its volume's outcome."""
+    engine = container_engine(container)
+    if engine is None:
+        return VolumeOutcome(database=False, dumped=False)
+    db_type, dump_tool = engine
+    dumped = backup_database(
+        container=container,
+        volume_dir=volume_dir,
+        db_type=db_type,
+        dump_tool=dump_tool,
+        databases_df=databases_df,
+        database_containers=database_containers,
+    )
+    return VolumeOutcome(database=True, dumped=dumped, engine=db_type)
+
+
+def _empty_databases_df() -> pd.DataFrame:
+    """
+    Create an empty DataFrame with the expected schema for databases.csv.
+
+    This allows the backup to continue without DB dumps when the CSV is missing
+    or empty (pandas EmptyDataError).
+    """
+    return pd.DataFrame(columns=list(COLUMNS))
+
+
+def load_databases_df(csv_path: str) -> pd.DataFrame:
+    """
+    Load databases.csv robustly.
+
+    - Missing file     -> warn, continue with empty df
+    - Empty file       -> warn, continue with empty df
+    - Valid CSV        -> return dataframe
+    """
+    try:
+        return pd.read_csv(csv_path, sep=DELIMITER, keep_default_na=False, dtype=str)
+    except FileNotFoundError:
+        print(
+            f"WARNING: databases.csv not found: {csv_path}. Continuing without database dumps.",
+            file=sys.stderr,
+            flush=True,
+        )
+        return _empty_databases_df()
+    except EmptyDataError:
+        print(
+            f"WARNING: databases.csv exists but is empty: {csv_path}. Continuing without database dumps.",
+            file=sys.stderr,
+            flush=True,
+        )
+        return _empty_databases_df()
+
+
+def backup_dumps_for_volume(
+    *,
+    containers: list[str],
+    vol_dir: str,
+    databases_df: pd.DataFrame,
+    database_containers: list[str],
+) -> VolumeOutcome:
+    """The volume's outcome across every container that mounts it."""
+    found_db = False
+    dumped_any = False
+    engine: str | None = None
+
+    for c in containers:
+        outcome = backup_mariadb_or_postgres(
+            container=c,
+            volume_dir=vol_dir,
+            databases_df=databases_df,
+            database_containers=database_containers,
+        )
+        if outcome.database:
+            found_db = True
+        if outcome.dumped:
+            dumped_any = True
+        if engine is None:
+            engine = outcome.engine
+
+    return VolumeOutcome(database=found_db, dumped=dumped_any, engine=engine)

src/baudolo/backup/layout.py

@@ -0,0 +1,61 @@
+"""Where a backup run puts its files, and how a finished run is stamped."""
+
+from __future__ import annotations
+
+import json
+import pathlib
+
+from dirval import create_stamp_file
+
+from baudolo.generation import MANIFEST_FILE, manifest_document
+
+from .shell import BackupError, execute_shell_command
+
+
+def get_machine_id() -> str:
+    return execute_shell_command(["sha256sum", "/etc/machine-id"])[0][0:64]
+
+
+def stamp_directory(version_dir: str) -> None:
+    """
+    Use dirval as a Python library to stamp the directory (no CLI dependency).
+    """
+    create_stamp_file(version_dir)
+
+
+def create_version_directory(versions_dir: str, backup_time: str) -> str:
+    version_dir = str(pathlib.Path(versions_dir) / backup_time)
+    try:
+        pathlib.Path(version_dir).mkdir(parents=True)
+    except FileExistsError:
+        raise BackupError(
+            f"generation {backup_time} already exists at {version_dir}; "
+            "another run claimed this second - refusing to write into it, "
+            "since rsync --delete would overwrite that generation"
+        ) from None
+    return version_dir
+
+
+def create_volume_directory(version_dir: str, volume_name: str) -> str:
+    path = pathlib.Path(version_dir) / volume_name
+    path.mkdir(parents=True, exist_ok=True)
+    return str(path)
+
+
+def write_manifest(version_dir: str, volumes: dict[str, dict[str, bool]]) -> str:
+    """Record the generation's layout and per-volume outcome.
+
+    Written before the directory is stamped, so the stamp covers it.
+
+    Args:
+        version_dir: the generation directory.
+        volumes: per volume name, ``database`` and ``dumped``.
+
+    Returns:
+        The path written.
+    """
+    path = pathlib.Path(version_dir) / MANIFEST_FILE
+    with path.open("w", encoding="utf-8") as handle:
+        json.dump(manifest_document(volumes), handle, indent=2, sort_keys=True)
+        handle.write("\n")
+    return str(path)

src/baudolo/backup/policy.py

@@ -0,0 +1,38 @@
+"""Which volumes are backed up, and which containers must stop for it."""
+
+from __future__ import annotations
+
+from .docker import get_image_info, is_swarm_task
+
+
+def is_image_ignored(container: str, images_no_backup_required: list[str]) -> bool:
+    if not images_no_backup_required:
+        return False
+    img = get_image_info(container)
+    return img in images_no_backup_required
+
+
+def volume_is_fully_ignored(
+    containers: list[str], images_no_backup_required: list[str]
+) -> bool:
+    """
+    Skip file backup only if all containers linked to the volume are ignored.
+    """
+    if not containers:
+        return False
+    return all(is_image_ignored(c, images_no_backup_required) for c in containers)
+
+
+def requires_stop(containers: list[str], images_no_stop_required: list[str]) -> bool:
+    """
+    Stop is required if ANY stoppable container image is NOT in the exact
+    image whitelist. Swarm task containers never count: baudolo must
+    not cycle them (see docker.is_swarm_task).
+    """
+    for c in containers:
+        if is_swarm_task(c):
+            continue
+        img = get_image_info(c)
+        if img not in images_no_stop_required:
+            return True
+    return False

src/baudolo/backup/shell.py

@@ -1,26 +1,75 @@
+"""Running external commands without a shell.
+
+Every command is an argv list. A database name, a password or a container name
+therefore cannot close a quote and start a second command, which a formatted
+string handed to ``shell=True`` allowed.
+"""
+
 from __future__ import annotations
 
+import os
 import subprocess
+from pathlib import Path
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from collections.abc import Mapping, Sequence
 
 
-class BackupException(Exception):
+class BackupError(Exception):
     """Generic exception for backup errors."""
 
 
-def execute_shell_command(command: str) -> list[str]:
-    """Execute a shell command and return its output lines."""
-    print(command, flush=True)
+def _child_env(env: Mapping[str, str] | None) -> dict[str, str] | None:
+    return None if env is None else {**os.environ, **env}
+
+
+def _fail(command: Sequence[str], returncode: int, out: bytes, err: bytes) -> None:
+    raise BackupError(
+        f"Error in command: {' '.join(command)}\n"
+        f"Output: {out}\nError: {err}\n"
+        f"Exit code: {returncode}"
+    )
+
+
+def execute_shell_command(
+    command: Sequence[str], *, env: Mapping[str, str] | None = None
+) -> list[str]:
+    """Run *command* and return its stdout lines.
+
+    Args:
+        command: argv, the program first.
+        env: variables added to the child's environment, for values that must
+            not appear in the argv of a process listing.
+    """
+    command = list(command)
+    print(" ".join(command), flush=True)
     process = subprocess.Popen(
-        [command],
-        stdout=subprocess.PIPE,
-        stderr=subprocess.PIPE,
-        shell=True,
+        command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, env=_child_env(env)
     )
     out, err = process.communicate()
     if process.returncode != 0:
-        raise BackupException(
-            f"Error in command: {command}\n"
-            f"Output: {out}\nError: {err}\n"
-            f"Exit code: {process.returncode}"
-        )
+        _fail(command, process.returncode, out, err)
     return [line.decode("utf-8") for line in out.splitlines()]
+
+
+def execute_to_file(
+    command: Sequence[str], out_file: str, *, env: Mapping[str, str] | None = None
+) -> None:
+    """Run *command*, writing its stdout to *out_file* only once it succeeded.
+
+    The output goes to a sibling temporary file first, so a partial or empty
+    stream from a failing dump never takes the place of a valid backup.
+    """
+    command = list(command)
+    print(" ".join(command), flush=True)
+    tmp = Path(f"{out_file}.tmp")
+    with tmp.open("wb") as handle:
+        process = subprocess.Popen(
+            command, stdout=handle, stderr=subprocess.PIPE, env=_child_env(env)
+        )
+        _, err = process.communicate()
+    if process.returncode != 0:
+        tmp.unlink()
+        _fail(command, process.returncode, b"", err)
+    tmp.replace(out_file)

src/baudolo/backup/snapshot.py

@@ -0,0 +1,176 @@
+"""Capture every volume from one atomic filesystem snapshot.
+
+Copying a live tree file by file cannot produce a point in time: a database can
+write between two files and leave a control file and its write-ahead log
+disagreeing, which no recovery can repair. A snapshot freezes the whole subject
+at once, so a database reads it as a crash and replays its log - a case it is
+built for. That also removes the reason to stop containers at all.
+
+The snapshot kind is stated by the caller rather than probed, because falling
+back to a live copy when a probe is inconclusive would hand out backups that
+look consistent and are not.
+
+Which volumes a snapshot of the subject contains is a different question, and
+it is decided per volume: a volume with a backing store of its own appears
+inside the snapshot as an existing empty directory, so copying from there
+succeeds and stores nothing. Such a volume is copied live instead - correct
+data without the point in time - while every other volume of the same run
+keeps its snapshot.
+"""
+
+from __future__ import annotations
+
+import os
+from contextlib import contextmanager
+from pathlib import Path
+from typing import TYPE_CHECKING
+
+from .shell import BackupError, execute_shell_command
+
+if TYPE_CHECKING:
+    from collections.abc import Callable, Iterator
+
+    from .volume import Backing
+
+KINDS = ("btrfs", "zfs")
+
+
+class SnapshotError(RuntimeError):
+    """A snapshot could not be created, resolved or removed."""
+
+
+def _resolver(subject: str, root: str) -> Callable[[str], str]:
+    def resolve(path: str) -> str:
+        # Exception: abspath, not Path.resolve() - resolve() follows symlinks,
+        # which would let a symlinked volume test as inside the subject.
+        relative = os.path.relpath(
+            os.path.abspath(path),  # noqa: PTH100
+            os.path.abspath(subject),  # noqa: PTH100
+        )
+        if relative.startswith(".."):
+            raise SnapshotError(f"{path} lies outside the snapshot subject {subject}")
+        resolved = root if relative == "." else str(Path(root) / relative)
+
+        # abspath drops a trailing separator, and rsync reads "dir/" as its
+        # contents where "dir" means the directory itself.
+        return resolved + os.sep if path.endswith(os.sep) else resolved
+
+    return resolve
+
+
+def _btrfs(
+    subject: str, name: str, run: Callable[[list[str]], list[str]]
+) -> tuple[str, list[str]]:
+    # The snapshot goes inside the subject, never beside it: the kernel rejects
+    # a snapshot whose destination is on another filesystem, which is exactly
+    # what the parent directory is when the subject is a mountpoint of its own.
+    target = str(Path(os.path.abspath(subject)) / f".{name}")  # noqa: PTH100 - see _resolver
+    run(["btrfs", "subvolume", "snapshot", "-r", subject, target])
+    return target, ["btrfs", "subvolume", "delete", target]
+
+
+def _zfs(
+    subject: str, name: str, run: Callable[[list[str]], list[str]]
+) -> tuple[str, list[str]]:
+    output = run(["zfs", "list", "-H", "-o", "name", subject])
+    dataset = (output[0] if output else "").strip()
+    if not dataset:
+        raise SnapshotError(f"no zfs dataset is mounted at {subject}")
+    run(["zfs", "snapshot", f"{dataset}@{name}"])
+    root = str(Path(subject) / ".zfs" / "snapshot" / name)
+    return root, ["zfs", "destroy", f"{dataset}@{name}"]
+
+
+_CREATE = {"btrfs": _btrfs, "zfs": _zfs}
+
+
+def unsnapshotted(backing: Backing, subject: str) -> str | None:
+    """Return why a snapshot of ``subject`` does not hold this volume's data.
+
+    Docker mounts a volume's own backing store lazily and unmounts it when the
+    last consumer stops, so the declaration is what gets checked: it is true at
+    every moment, where the mount table is only true while a container happens
+    to hold the volume.
+
+    Args:
+        backing: the volume as the daemon describes it.
+        subject: the snapshot subject, e.g. ``/var/lib/docker``.
+
+    Returns:
+        The reason, or None when the snapshot holds the volume.
+    """
+    if backing.driver != "local":
+        return f"it uses the {backing.driver} driver"
+    if backing.options:
+        return f"it declares its own backing store {backing.options}"
+    if not backing.mountpoint:
+        return "it reports no mountpoint"
+    real = os.path.realpath(backing.mountpoint)
+    if os.path.ismount(real):
+        return f"its mountpoint {backing.mountpoint} sits on its own mount"
+    try:
+        crosses = (
+            Path(real).stat().st_dev != Path(os.path.realpath(subject)).stat().st_dev
+        )
+    except OSError as error:
+        return f"its mountpoint {backing.mountpoint} could not be read: {error}"
+    if crosses:
+        return f"its mountpoint {backing.mountpoint} crosses a filesystem boundary"
+    return None
+
+
+def snapshot_source(
+    resolve: Callable[[str], str], backing: Backing, subject: str
+) -> tuple[str | None, str]:
+    """Resolve where to read a volume from, and why if not from the snapshot.
+
+    Returns:
+        ``(path, "")`` to copy from the snapshot, or ``(None, reason)`` to copy
+        it live.
+    """
+    reason = unsnapshotted(backing, subject)
+    if reason:
+        return None, reason
+    try:
+        source = resolve(backing.source)
+    except SnapshotError as error:
+        return None, str(error)
+    if not Path(source).is_dir():
+        return None, "it was created after the snapshot was taken"
+    return source, ""
+
+
+@contextmanager
+def volume_snapshot(
+    kind: str,
+    subject: str,
+    tag: str,
+    run: Callable[[list[str]], list[str]] = execute_shell_command,
+) -> Iterator[Callable[[str], str]]:
+    """Yield a resolver mapping a path under ``subject`` into a snapshot of it.
+
+    Args:
+        kind: ``btrfs`` or ``zfs``; the caller states it, nothing is probed.
+        subject: the btrfs subvolume or zfs dataset mountpoint holding the
+            volumes, e.g. ``/var/lib/docker``.
+        tag: unique suffix for the snapshot name, e.g. the backup timestamp.
+        run: shell runner, injected so the mechanics are testable.
+
+    Raises:
+        SnapshotError: the kind is unknown, or the snapshot cannot be created.
+            Removal failure is reported, not raised: a leftover snapshot is a
+            cleanup problem and must not discard a generation that is complete.
+    """
+    create = _CREATE.get(kind)
+    if create is None:
+        raise SnapshotError(f"unknown snapshot kind {kind!r}; expected one of {KINDS}")
+
+    root, remove = create(subject, f"baudolo-{tag}", run)
+    try:
+        yield _resolver(subject, root)
+    finally:
+        try:
+            run(remove)
+        except BackupError as error:
+            # Raising here would also mask whatever the body raised.
+            print(f"WARNING: {root} could not be removed: {error}", flush=True)

src/baudolo/backup/volume.py

@@ -1,42 +1,93 @@
 from __future__ import annotations
 
+import json
 import os
 import pathlib
+from dataclasses import dataclass, field
 
-from .shell import BackupException, execute_shell_command
+from baudolo.generation import FILES_DIR
+
+from .shell import BackupError, execute_shell_command
 
 
-def get_storage_path(volume_name: str) -> str:
-    path = execute_shell_command(
-        f"docker volume inspect --format '{{{{ .Mountpoint }}}}' {volume_name}"
+@dataclass(frozen=True)
+class Backing:
+    """Where a docker volume actually keeps its data.
+
+    Args:
+        mountpoint: the path the daemon reports.
+        driver: the volume driver, ``local`` for the built-in one.
+        options: the driver options; a non-empty map means the mountpoint is a
+            mount target rather than the storage itself.
+    """
+
+    mountpoint: str
+    driver: str = "local"
+    options: dict = field(default_factory=dict)
+
+    @property
+    def source(self) -> str:
+        return f"{self.mountpoint}/"
+
+
+def inspect_backing(volume_name: str) -> Backing:
+    reported = execute_shell_command(
+        ["docker", "volume", "inspect", "--format", "{{json .}}", volume_name]
     )[0]
-    return f"{path}/"
+    data = json.loads(reported)
+    return Backing(
+        data.get("Mountpoint") or "",
+        data.get("Driver") or "",
+        data.get("Options") or {},
+    )
 
 
-def get_last_backup_dir(versions_dir: str, volume_name: str, current_backup_dir: str) -> str | None:
+def get_last_backup_dir(
+    versions_dir: str, volume_name: str, current_backup_dir: str
+) -> str | None:
     versions = sorted(os.listdir(versions_dir), reverse=True)
     for version in versions:
-        candidate = os.path.join(versions_dir, version, volume_name, "files", "")
-        if candidate != current_backup_dir and os.path.isdir(candidate):
+        candidate = f"{pathlib.Path(versions_dir) / version / volume_name / FILES_DIR}/"
+        if candidate != current_backup_dir and pathlib.Path(candidate).is_dir():
             return candidate
     return None
 
 
-def backup_volume(versions_dir: str, volume_name: str, volume_dir: str) -> None:
-    """Perform incremental file backup of a Docker volume."""
-    dest = os.path.join(volume_dir, "files") + "/"
+def backup_volume(
+    versions_dir: str,
+    volume_name: str,
+    volume_dir: str,
+    *,
+    authoritative: bool,
+    source: str,
+) -> None:
+    """Perform incremental file backup of a Docker volume.
+
+    Args:
+        authoritative: compare source and destination by content instead of by
+            size and whole-second mtime. Required on a pass whose destination was
+            already written from a live source, where a file can differ while
+            both attributes still agree.
+        source: directory to read from - the volume's mountpoint, or its path
+            inside a snapshot.
+    """
+    dest = f"{pathlib.Path(volume_dir) / FILES_DIR}/"
     pathlib.Path(dest).mkdir(parents=True, exist_ok=True)
 
     last = get_last_backup_dir(versions_dir, volume_name, dest)
-    link_dest = f"--link-dest='{last}'" if last else ""
-    source = get_storage_path(volume_name)
-
-    cmd = f"rsync -abP --delete --delete-excluded {link_dest} {source} {dest}"
+    cmd = ["rsync", "-aP", "--no-D", "--delete", "--delete-excluded"]
+    if authoritative:
+        cmd.append("--checksum")
+    if last:
+        cmd.append(f"--link-dest={last}")
+    cmd += [source, dest]
 
     try:
         execute_shell_command(cmd)
-    except BackupException as e:
+    except BackupError as e:
         if "file has vanished" in str(e):
-            print("Warning: Some files vanished before transfer. Continuing.", flush=True)
+            print(
+                "Warning: Some files vanished before transfer. Continuing.", flush=True
+            )
         else:
             raise

src/baudolo/databases.py

@@ -0,0 +1,103 @@
+"""The databases.csv contract: its columns, its delimiter, and what a row means.
+
+``baudolo-seed`` writes the file, the backup reads it to learn which dumps to
+take, and a restore consumer reads it again to replay them. Stating the schema
+once keeps a column or a convention added here from being invisible to the
+other two.
+
+Field values are handed back exactly as they stand in the file. A password may
+legitimately begin or end with a space, so stripping belongs to the caller that
+compares, never to the reader.
+"""
+
+from __future__ import annotations
+
+import csv
+import re
+from pathlib import Path
+from typing import NamedTuple
+
+COLUMNS = ("instance", "database", "username", "password")
+DELIMITER = ";"
+CLUSTER_ROW = "*"
+
+_NAME_RE = re.compile(r"^[a-zA-Z0-9_][a-zA-Z0-9_-]*$")
+
+
+class DatabasesCsvError(ValueError):
+    """A row does not match the contract."""
+
+
+class Row(NamedTuple):
+    """One databases.csv row, verbatim.
+
+    ``database`` holds :data:`CLUSTER_ROW` when the whole instance is dumped.
+    """
+
+    instance: str
+    database: str
+    username: str
+    password: str
+
+    @property
+    def is_cluster(self) -> bool:
+        return self.database.strip() == CLUSTER_ROW
+
+
+def validate_database(value: str | None, *, instance: str) -> str:
+    """The database column of one row, or raise.
+
+    The name reaches a shell as part of the dump command, so it is checked
+    where it is read as well as where it is written: a file edited by hand
+    never passed the seed.
+
+    Args:
+        value: the raw column.
+        instance: named in the error, so a bad row can be found.
+
+    Raises:
+        DatabasesCsvError: the column is empty, literally ``nan``, or holds
+            anything but letters, numbers, ``_`` and ``-``.
+    """
+    text = (value or "").strip()
+    if not text:
+        raise DatabasesCsvError(
+            f"Invalid databases.csv entry for instance '{instance}': column "
+            f"'database' must be '{CLUSTER_ROW}' or a concrete database name "
+            "(not empty)."
+        )
+    if text == CLUSTER_ROW:
+        return CLUSTER_ROW
+    if text.lower() == "nan":
+        raise DatabasesCsvError(
+            f"Invalid databases.csv entry for instance '{instance}': "
+            "database must not be 'nan'."
+        )
+    if not _NAME_RE.match(text):
+        raise DatabasesCsvError(
+            f"Invalid databases.csv entry for instance '{instance}': invalid "
+            f"database name '{text}'. Allowed: letters, numbers, '_' and '-'."
+        )
+    return text
+
+
+def read_rows(csv_path: str) -> list[Row]:
+    """Every row of the file in file order, header skipped, blank rows dropped.
+
+    Raises:
+        DatabasesCsvError: a row holds fewer columns than :data:`COLUMNS`.
+    """
+    rows: list[Row] = []
+    with Path(csv_path).open(newline="", encoding="utf-8") as handle:
+        reader = csv.reader(handle, delimiter=DELIMITER)
+        next(reader, None)
+        for raw in reader:
+            if not any(field.strip() for field in raw):
+                continue
+            if len(raw) < len(COLUMNS):
+                raise DatabasesCsvError(
+                    f"{csv_path} has a row with {len(raw)} column(s), "
+                    f"expected {len(COLUMNS)}"
+                )
+            rows.append(Row(*raw[: len(COLUMNS)]))
+    return rows

src/baudolo/generation.py

@@ -0,0 +1,54 @@
+"""The on-disk shape of a generation, and the manifest that states it.
+
+Every name a reader needs to find payload in a generation is declared here
+once, and written into each generation's own manifest. A consumer therefore
+never has to hardcode the layout or match this package's version: it reads
+what the run that produced the tree recorded.
+
+The manifest also carries what only the run itself can know: per volume,
+``database`` (it held one), ``dumped`` (a dump was produced for it) and
+``engine`` (which one was detected). Both flags true is a replayable dump;
+``database`` without ``dumped`` is a raw copy of live engine files.
+
+Kept import-free: consumers read the manifest with nothing but ``json``, on
+hosts that do not have this package installed.
+"""
+
+from __future__ import annotations
+
+FILES_DIR = "files"
+SQL_DIR = "sql"
+DUMP_SUFFIX = ".backup.sql"
+CLUSTER_SUFFIX = ".cluster.backup.sql"
+
+MANIFEST_FILE = "manifest.json"
+MANIFEST_SCHEMA = 1
+
+
+def manifest_document(volumes: dict[str, object]) -> dict[str, object]:
+    """The manifest a finished run writes.
+
+    Args:
+        volumes: per volume name, an object carrying ``database``, ``dumped``
+            and ``engine`` -- a ``baudolo.backup.dumps.VolumeOutcome``.
+
+    Returns:
+        The document, ready for ``json.dump``.
+    """
+    return {
+        "schema": MANIFEST_SCHEMA,
+        "layout": {
+            "files_dir": FILES_DIR,
+            "sql_dir": SQL_DIR,
+            "dump_suffix": DUMP_SUFFIX,
+            "cluster_suffix": CLUSTER_SUFFIX,
+        },
+        "volumes": {
+            name: {
+                "database": bool(outcome.database),
+                "dumped": bool(outcome.dumped),
+                "engine": outcome.engine,
+            }
+            for name, outcome in sorted(volumes.items())
+        },
+    }

src/baudolo/restore/__init__.py

@@ -1 +1 @@
-__all__ = ["main"]
+__all__ = ["main"]

src/baudolo/restore/__main__.py

@@ -3,10 +3,11 @@ from __future__ import annotations
 import argparse
 import sys
 
-from .paths import BackupPaths
-from .files import restore_volume_files
-from .db.postgres import restore_postgres_sql
+from .db.cluster import restore_cluster_sql
 from .db.mariadb import restore_mariadb_sql
+from .db.postgres import restore_postgres_sql
+from .files import restore_volume_files
+from .paths import BackupPaths
 
 
 def _add_common_backup_args(p: argparse.ArgumentParser) -> None:
@@ -21,8 +22,22 @@ def _add_common_backup_args(p: argparse.ArgumentParser) -> None:
     )
     p.add_argument(
         "--repo-name",
-        default="backup-docker-to-local",
-        help="Backup repo folder name under <backups-dir>/<hash>/ (default: backup-docker-to-local)",
+        required=True,
+        help="Backup repo folder name under <backups-dir>/<hash>/",
+    )
+
+
+def _add_common_engine_args(p: argparse.ArgumentParser) -> None:
+    p.add_argument("--container", required=True)
+    p.add_argument("--db-password", required=True)
+    p.add_argument("--empty", action="store_true")
+    p.add_argument(
+        "--no-version-check",
+        action="store_true",
+        help=(
+            "Replay even if the dump comes from a newer engine than the target. "
+            "With --empty this can leave an emptied database behind."
+        ),
     )
 
 
@@ -33,15 +48,8 @@ def main(argv: list[str] | None = None) -> int:
     )
     sub = parser.add_subparsers(dest="cmd", required=True)
 
-    # ------------------------------------------------------------------
-    # files
-    # ------------------------------------------------------------------
     p_files = sub.add_parser("files", help="Restore files into a docker volume")
     _add_common_backup_args(p_files)
-    p_files.add_argument(
-        "--rsync-image",
-        default="ghcr.io/kevinveenbirkenbach/alpine-rsync",
-    )
     p_files.add_argument(
         "--source-volume",
         default=None,
@@ -52,34 +60,40 @@ def main(argv: list[str] | None = None) -> int:
         ),
     )
 
-    # ------------------------------------------------------------------
-    # postgres
-    # ------------------------------------------------------------------
     p_pg = sub.add_parser("postgres", help="Restore a single PostgreSQL database dump")
     _add_common_backup_args(p_pg)
-    p_pg.add_argument("--container", required=True)
+    _add_common_engine_args(p_pg)
     p_pg.add_argument("--db-name", required=True)
     p_pg.add_argument("--db-user", default=None, help="Defaults to db-name if omitted")
-    p_pg.add_argument("--db-password", required=True)
-    p_pg.add_argument("--empty", action="store_true")
 
-    # ------------------------------------------------------------------
-    # mariadb
-    # ------------------------------------------------------------------
-    p_mdb = sub.add_parser("mariadb", help="Restore a single MariaDB/MySQL-compatible dump")
+    p_cluster = sub.add_parser(
+        "cluster", help="Restore a full PostgreSQL cluster dump (pg_dumpall)"
+    )
+    _add_common_backup_args(p_cluster)
+    _add_common_engine_args(p_cluster)
+    p_cluster.add_argument(
+        "--instance",
+        required=True,
+        help="Instance the dump was taken from; names <instance>.cluster.backup.sql",
+    )
+    p_cluster.add_argument(
+        "--db-user",
+        required=True,
+        help="Superuser of the instance; the dump creates roles and databases",
+    )
+
+    p_mdb = sub.add_parser(
+        "mariadb", help="Restore a single MariaDB/MySQL-compatible dump"
+    )
     _add_common_backup_args(p_mdb)
-    p_mdb.add_argument("--container", required=True)
+    _add_common_engine_args(p_mdb)
     p_mdb.add_argument("--db-name", required=True)
     p_mdb.add_argument("--db-user", default=None, help="Defaults to db-name if omitted")
-    p_mdb.add_argument("--db-password", required=True)
-    p_mdb.add_argument("--empty", action="store_true")
 
     args = parser.parse_args(argv)
 
     try:
         if args.cmd == "files":
-            # target volume = args.volume_name
-            # source volume (backup key) defaults to target volume
             source_volume = args.source_volume or args.volume_name
 
             bp_files = BackupPaths(
@@ -93,7 +107,6 @@ def main(argv: list[str] | None = None) -> int:
             return restore_volume_files(
                 args.volume_name,
                 bp_files.files_dir(),
-                rsync_image=args.rsync_image,
             )
 
         if args.cmd == "postgres":
@@ -111,6 +124,24 @@ def main(argv: list[str] | None = None) -> int:
                     backups_dir=args.backups_dir,
                 ).sql_file(args.db_name),
                 empty=args.empty,
+                check_version=not args.no_version_check,
+            )
+            return 0
+
+        if args.cmd == "cluster":
+            restore_cluster_sql(
+                container=args.container,
+                user=args.db_user,
+                password=args.db_password,
+                sql_path=BackupPaths(
+                    args.volume_name,
+                    args.backup_hash,
+                    args.version,
+                    repo_name=args.repo_name,
+                    backups_dir=args.backups_dir,
+                ).cluster_file(args.instance),
+                empty=args.empty,
+                check_version=not args.no_version_check,
             )
             return 0
 
@@ -129,13 +160,14 @@ def main(argv: list[str] | None = None) -> int:
                     backups_dir=args.backups_dir,
                 ).sql_file(args.db_name),
                 empty=args.empty,
+                check_version=not args.no_version_check,
             )
             return 0
 
         parser.error("Unhandled command")
-        return 2
+        return 2  # noqa: TRY300 - the try wraps the whole dispatch on purpose
 
-    except Exception as e:
+    except Exception as e:  # noqa: BLE001 - CLI boundary: any failure becomes exit 1
         print(f"ERROR: {e}", file=sys.stderr)
         return 1
 

src/baudolo/restore/db/__init__.py

@@ -1 +1 @@
-"""Database restore handlers (Postgres, MariaDB/MySQL)."""
+"""Database restore handlers (Postgres, MariaDB/MySQL)."""

src/baudolo/restore/db/cluster.py

@@ -0,0 +1,252 @@
+"""Replay a full PostgreSQL cluster dump produced by ``pg_dumpall``.
+
+The backup side writes one when a databases.csv row asks for every database of
+an instance (``database = '*'``, see ``backup/db.py``). Until now nothing read
+it back, so that dump was stored and unrestorable - a format whose producer has
+no consumer.
+
+A cluster stream differs from a single-database one in three ways that decide
+the implementation:
+
+* it recreates roles and databases, so it must be replayed against the control
+  database rather than into a target database;
+* ``CREATE DATABASE`` cannot run inside a transaction block, so unlike
+  :mod:`baudolo.restore.db.postgres` the replay must not be wrapped in
+  ``--single-transaction``;
+* it is replayed as a superuser, so the superuser-only statements that the
+  single-database path filters out are exactly the ones that have to survive.
+"""
+
+from __future__ import annotations
+
+import re
+import tempfile
+from pathlib import Path
+from typing import TYPE_CHECKING
+
+from baudolo.restore.run import docker_exec
+
+from .version import guard
+
+if TYPE_CHECKING:
+    from collections.abc import Iterable, Iterator
+
+CONTROL_DB = "postgres"
+_CLUSTER_PRECLEAN_SQL = Path(__file__).parent / "cluster_preclean.sql"
+_CREATE_ROLE = re.compile(rb'^CREATE ROLE "?([^";]+)"?;\s*$')
+_CREATE_DATABASE = re.compile(rb"^CREATE DATABASE\s+(.*)$")
+_CREATE_ROLE_LINE = re.compile(rb"^CREATE ROLE\s+(.*)$")
+_CONNECT = re.compile(rb"^\\connect\s+(.*)$")
+_NO_ROWS = "SELECT ''::text WHERE false"
+
+
+def _first_identifier(rest: str) -> str | None:
+    """The first SQL identifier in *rest*, quoted or bare.
+
+    A quoted identifier may hold spaces and doubled quotes, so it cannot be
+    read with a character class that stops at whitespace - which is how a
+    database called ``odd name`` used to leave the inventory as ``odd``.
+    """
+    text = rest.strip()
+    if not text:
+        return None
+    if text.startswith('"'):
+        out = []
+        index = 1
+        while index < len(text):
+            char = text[index]
+            if char == '"':
+                if index + 1 < len(text) and text[index + 1] == '"':
+                    out.append('"')
+                    index += 2
+                    continue
+                return "".join(out)
+            out.append(char)
+            index += 1
+        return None
+    return re.split(r"[\s;(]", text, maxsplit=1)[0] or None
+
+
+def _connect_target(rest: str) -> str | None:
+    """The database a ``\\connect`` line switches to.
+
+    psql options precede the name (``\\connect -reuse-previous=on dbname=x``),
+    and the name may arrive as a ``dbname=`` assignment rather than bare.
+    """
+    for token in rest.strip().split():
+        if token.startswith("-"):
+            continue
+        if token.startswith("dbname="):
+            return _first_identifier(token[len("dbname=") :])
+        return _first_identifier(rest.strip()[rest.strip().index(token) :])
+    return None
+
+
+def dump_inventory(sql_path: str) -> tuple[list[str], list[str]]:
+    """The databases and roles a cluster dump recreates.
+
+    Args:
+        sql_path: the ``pg_dumpall`` stream.
+
+    Returns:
+        ``(databases, roles)``, each in the order the dump names them. The
+        pre-clean is scoped to these: everything else in the instance belongs
+        to no backup this restore holds, and dropping it would destroy data
+        the replay cannot bring back.
+    """
+    databases: list[str] = []
+    roles: list[str] = []
+    with Path(sql_path).open("rb") as handle:
+        for raw in handle:
+            line = raw.decode("utf-8", "replace")
+            for pattern, sink, read in (
+                (_CREATE_DATABASE, databases, _first_identifier),
+                (_CONNECT, databases, _connect_target),
+                (_CREATE_ROLE_LINE, roles, _first_identifier),
+            ):
+                found = pattern.match(raw)
+                if not found:
+                    continue
+                name = read(line[found.start(1) :])
+                if name and name not in sink:
+                    sink.append(name)
+    return databases, roles
+
+
+def preclean_sql() -> str:
+    """The catalog-wide pre-clean, safe only behind the instance check."""
+    with _CLUSTER_PRECLEAN_SQL.open(encoding="utf-8") as preclean:
+        return preclean.read()
+
+
+def instance_databases(container: str, user: str, docker_env: dict) -> list[str]:
+    """The instance's own databases, templates and control database aside."""
+    listed = docker_exec(
+        container,
+        [
+            "psql",
+            "-U",
+            user,
+            "-d",
+            CONTROL_DB,
+            "-tAc",
+            (
+                "SELECT datname FROM pg_database "
+                "WHERE NOT datistemplate AND datname <> current_database()"
+            ),
+        ],
+        capture=True,
+        docker_env=docker_env,
+    ).stdout
+    text = listed.decode() if isinstance(listed, bytes) else listed
+    return [name for name in text.split() if name]
+
+
+def assert_instance_matches_dump(
+    container: str, user: str, sql_path: str, docker_env: dict
+) -> None:
+    """Refuse ``--empty`` on an instance holding anything the dump lacks.
+
+    The pre-clean is a catalog-wide sweep, so a foreign database would be
+    destroyed with no way back. Scoping the sweep instead is not a fix: a
+    surviving database that owns or grants to one of the dump's roles pins
+    that role in pg_shdepend, and DROP ROLE then fails after the dump's own
+    databases are already gone.
+
+    Raises:
+        RuntimeError: the instance carries databases this dump cannot restore.
+    """
+    dumped, _roles = dump_inventory(sql_path)
+    present = instance_databases(container, user, docker_env)
+    foreign = sorted(set(present) - set(dumped))
+    if foreign:
+        raise RuntimeError(
+            f"{container} also holds {', '.join(foreign)}, which "
+            f"{Path(sql_path).name} does not carry. --empty wipes the "
+            "instance, so those would be destroyed with nothing to restore "
+            "them from. Move them off this instance, or drop them yourself if "
+            "they are disposable."
+        )
+
+
+def _psql(user: str) -> list[str]:
+    """The replay client: no --single-transaction, CREATE DATABASE forbids it."""
+    return ["psql", "-v", "ON_ERROR_STOP=1", "-U", user, "-d", CONTROL_DB]
+
+
+def filter_own_role_creation(lines: Iterable[bytes], user: str) -> Iterator[bytes]:
+    """Drop the ``CREATE ROLE`` of the role holding this session.
+
+    A pg_dumpall stream recreates every role of the cluster, the bootstrap
+    superuser included, and the pre-clean cannot drop the one it is connected
+    as - so that single statement always collides. Its ``ALTER ROLE`` is kept:
+    that is what re-applies the attributes and the password the dump captured.
+
+    Args:
+        lines: dump lines including their trailing newlines.
+        user: the connecting role.
+
+    Yields:
+        Every line except that one CREATE.
+    """
+    for line in lines:
+        found = _CREATE_ROLE.match(line)
+        if found and found.group(1).decode() == user:
+            continue
+        yield line
+
+
+def restore_cluster_sql(
+    *,
+    container: str,
+    user: str,
+    password: str,
+    sql_path: str,
+    empty: bool,
+    check_version: bool = True,
+) -> None:
+    """Replay a pg_dumpall stream into a running instance.
+
+    Args:
+        container: the running engine to replay into.
+        user: a superuser of that instance; the dump creates roles and
+            databases, which an application role may not do.
+        password: its password, handed to psql through the container's env.
+        sql_path: the ``<instance>.cluster.backup.sql`` of a generation.
+        empty: drop the cluster's databases and roles first. Without it the
+            replay stops at the first object that already exists, which is the
+            honest outcome: recreating a cluster over a populated one is a
+            decision, not a default.
+        check_version: refuse a dump from a newer major version than the
+            running engine before anything is dropped.
+    """
+    if not Path(sql_path).is_file():
+        raise FileNotFoundError(sql_path)
+
+    if check_version:
+        guard(
+            sql_path=sql_path,
+            engine="postgres",
+            container=container,
+            user=user,
+            password=password,
+        )
+
+    docker_env = {"PGPASSWORD": password}
+
+    if empty:
+        assert_instance_matches_dump(container, user, sql_path, docker_env)
+        docker_exec(
+            container,
+            _psql(user),
+            stdin=preclean_sql().encode(),
+            docker_env=docker_env,
+        )
+
+    with Path(sql_path).open("rb") as src, tempfile.TemporaryFile() as filtered:
+        for line in filter_own_role_creation(src, user):
+            filtered.write(line)
+        filtered.seek(0)
+        docker_exec(container, _psql(user), stdin=filtered, docker_env=docker_env)
+
+    print(f"PostgreSQL cluster restore complete from '{Path(sql_path).name}'.")

src/baudolo/restore/db/cluster_preclean.sql

@@ -0,0 +1,33 @@
+-- Pre-clean for `restore cluster --empty`. A pg_dumpall stream recreates roles
+-- and databases, so replaying it into a populated cluster dies on the first
+-- CREATE ROLE. Emitted as one DROP per row and run via \gexec so each executes
+-- as its own top-level statement: DROP DATABASE cannot run inside a
+-- transaction block, which rules out a single DO block.
+-- The phase column pins the order: databases must be gone before their owners
+-- can be dropped, and DROP OWNED BY releases what a role still holds in the
+-- control database. Template databases, the control database itself, the pg_*
+-- system roles and the connecting role are kept - the dump does not recreate
+-- them and dropping them would end the session.
+-- The sweep stays catalog-wide on purpose: a scoped one leaves databases that
+-- pin a dumped role in pg_shdepend, and phase 3 then fails after phase 1 has
+-- already dropped. assert_instance_matches_dump refuses before this runs.
+SELECT statement
+  FROM (
+    SELECT 1 AS phase,
+           format('DROP DATABASE IF EXISTS %I', datname) AS statement
+      FROM pg_database
+     WHERE NOT datistemplate
+       AND datname <> current_database()
+    UNION ALL
+    SELECT 2, format('DROP OWNED BY %I', rolname)
+      FROM pg_roles
+     WHERE NOT starts_with(rolname, 'pg_')
+       AND rolname <> current_user
+    UNION ALL
+    SELECT 3, format('DROP ROLE IF EXISTS %I', rolname)
+      FROM pg_roles
+     WHERE NOT starts_with(rolname, 'pg_')
+       AND rolname <> current_user
+  ) drops
+ ORDER BY phase
+\gexec

src/baudolo/restore/db/empty_preclean.sql

@@ -0,0 +1,64 @@
+-- Owner-filtered pre-clean for `restore --empty`. Emitted as one DROP per row and
+-- run via \gexec so each executes as its own top-level statement: a single DO-block
+-- would run every DROP in one transaction and exhaust max_locks_per_transaction on
+-- large schemas (e.g. gitlab). Also drops user-owned non-public schemas so a dump
+-- that CREATE SCHEMAs (e.g. discourse's discourse_functions) does not fail on an
+-- already-existing schema. Extension members (pg_trgm's set_limit) are
+-- superuser-owned; IF EXISTS absorbs the CASCADE fallout.
+SELECT format('DROP %s IF EXISTS public.%s CASCADE', obj.type, obj.name)
+  FROM (
+    SELECT format('%I', c.relname) AS name,
+           CASE c.relkind
+             WHEN 'v' THEN 'VIEW'
+             WHEN 'm' THEN 'MATERIALIZED VIEW'
+             WHEN 'f' THEN 'FOREIGN TABLE'
+             ELSE 'TABLE'
+           END AS type
+      FROM pg_class c JOIN pg_namespace n ON n.oid = c.relnamespace
+     WHERE n.nspname = 'public' AND c.relkind IN ('r', 'p', 'v', 'm', 'f')
+       AND pg_get_userbyid(c.relowner) = current_user
+    UNION ALL
+    -- Overloaded functions share a proname; DROP needs the identity
+    -- signature or psql aborts with "function name is not unique".
+    SELECT format('%I(%s)', p.proname, pg_get_function_identity_arguments(p.oid)) AS name,
+           CASE p.prokind WHEN 'p' THEN 'PROCEDURE' ELSE 'FUNCTION' END AS type
+      FROM pg_proc p JOIN pg_namespace n ON n.oid = p.pronamespace
+     WHERE n.nspname = 'public' AND p.prokind IN ('f', 'p', 'w')
+       AND pg_get_userbyid(p.proowner) = current_user
+    UNION ALL
+    SELECT format('%I', c.relname) AS name, 'SEQUENCE' AS type
+      FROM pg_class c JOIN pg_namespace n ON n.oid = c.relnamespace
+     WHERE n.nspname = 'public' AND c.relkind = 'S'
+       AND pg_get_userbyid(c.relowner) = current_user
+    UNION ALL
+    SELECT format('%I', t.typname) AS name, 'TYPE' AS type
+      FROM pg_type t JOIN pg_namespace n ON n.oid = t.typnamespace
+     WHERE n.nspname = 'public'
+       AND pg_get_userbyid(t.typowner) = current_user
+       AND (t.typtype IN ('e', 'd')
+            OR (t.typtype = 'c' AND EXISTS (
+                  SELECT 1 FROM pg_class c2
+                   WHERE c2.oid = t.typrelid AND c2.relkind = 'c')))
+    UNION ALL
+    SELECT format('%I', col.collname) AS name, 'COLLATION' AS type
+      FROM pg_collation col JOIN pg_namespace n ON n.oid = col.collnamespace
+     WHERE n.nspname = 'public'
+       AND pg_get_userbyid(col.collowner) = current_user
+    UNION ALL
+    SELECT format('%I', ts.cfgname) AS name, 'TEXT SEARCH CONFIGURATION' AS type
+      FROM pg_ts_config ts JOIN pg_namespace n ON n.oid = ts.cfgnamespace
+     WHERE n.nspname = 'public'
+       AND pg_get_userbyid(ts.cfgowner) = current_user
+    UNION ALL
+    SELECT format('%I', d.dictname) AS name, 'TEXT SEARCH DICTIONARY' AS type
+      FROM pg_ts_dict d JOIN pg_namespace n ON n.oid = d.dictnamespace
+     WHERE n.nspname = 'public'
+       AND pg_get_userbyid(d.dictowner) = current_user
+  ) obj
+UNION ALL
+SELECT format('DROP SCHEMA IF EXISTS %I CASCADE', n.nspname)
+  FROM pg_namespace n
+ WHERE NOT starts_with(n.nspname, 'pg_')
+   AND n.nspname NOT IN ('public', 'information_schema')
+   AND pg_get_userbyid(n.nspowner) = current_user
+\gexec

src/baudolo/restore/db/mariadb.py

@@ -1,9 +1,13 @@
 from __future__ import annotations
 
-import os
 import sys
+from pathlib import Path
 
-from ..run import docker_exec, docker_exec_sh
+from baudolo.restore.run import docker_exec, docker_exec_sh
+
+from .version import guard
+
+_NO_CLIENT = "ERROR: neither 'mariadb' nor 'mysql' found in container."
 
 
 def _pick_client(container: str) -> str:
@@ -19,12 +23,13 @@ exit 42
 """
     try:
         out = docker_exec_sh(container, script, capture=True).stdout.decode().strip()
-        if not out:
-            raise RuntimeError("empty client detection output")
-        return out
-    except Exception as e:
-        print("ERROR: neither 'mariadb' nor 'mysql' found in container.", file=sys.stderr)
-        raise e
+    except Exception:
+        print(_NO_CLIENT, file=sys.stderr)
+        raise
+    if not out:
+        print(_NO_CLIENT, file=sys.stderr)
+        raise RuntimeError("empty client detection output")
+    return out
 
 
 def restore_mariadb_sql(
@@ -35,21 +40,25 @@ def restore_mariadb_sql(
     password: str,
     sql_path: str,
     empty: bool,
+    check_version: bool = True,
 ) -> None:
     client = _pick_client(container)
 
-    if not os.path.isfile(sql_path):
+    if not Path(sql_path).is_file():
         raise FileNotFoundError(sql_path)
 
-    if empty:
-        # IMPORTANT:
-        # Do NOT hardcode 'mysql' here. Use the detected client.
-        # MariaDB 11 images may not contain the mysql binary at all.
-        docker_exec(
-            container,
-            [client, "-u", user, f"--password={password}", "-e", "SET FOREIGN_KEY_CHECKS=0;"],
+    if check_version:
+        guard(
+            sql_path=sql_path,
+            engine="mariadb",
+            container=container,
+            user=user,
+            password=password,
+            client=client,
         )
 
+    if empty:
+        # Do not hardcode 'mysql': MariaDB 11 images may not ship that binary.
         result = docker_exec(
             container,
             [
@@ -59,13 +68,22 @@ def restore_mariadb_sql(
                 f"--password={password}",
                 "-N",
                 "-e",
-                f"SELECT table_name FROM information_schema.tables WHERE table_schema = '{db_name}';",
+                f"SELECT table_name FROM information_schema.tables WHERE table_schema = '{db_name}';",  # noqa: S608 - validate_database() constrains the name to ^[a-zA-Z0-9_][a-zA-Z0-9_-]*$
             ],
             capture=True,
         )
         tables = result.stdout.decode().split()
 
-        for tbl in tables:
+        if tables:
+            # SET FOREIGN_KEY_CHECKS is session-scoped, so it must share one
+            # client session with the DROPs or FK constraints still fire.
+            drop_sql = (
+                "SET FOREIGN_KEY_CHECKS=0; "
+                + " ".join(
+                    f"DROP TABLE IF EXISTS `{db_name}`.`{tbl}`;" for tbl in tables
+                )
+                + " SET FOREIGN_KEY_CHECKS=1;"
+            )
             docker_exec(
                 container,
                 [
@@ -74,16 +92,13 @@ def restore_mariadb_sql(
                     user,
                     f"--password={password}",
                     "-e",
-                    f"DROP TABLE IF EXISTS `{db_name}`.`{tbl}`;",
+                    drop_sql,
                 ],
             )
 
+    with Path(sql_path).open("rb") as f:
         docker_exec(
-            container,
-            [client, "-u", user, f"--password={password}", "-e", "SET FOREIGN_KEY_CHECKS=1;"],
+            container, [client, "-u", user, f"--password={password}", db_name], stdin=f
         )
 
-    with open(sql_path, "rb") as f:
-        docker_exec(container, [client, "-u", user, f"--password={password}", db_name], stdin=f)
-
     print(f"MariaDB/MySQL restore complete for db '{db_name}'.")

src/baudolo/restore/db/postgres.py

@@ -1,8 +1,46 @@
 from __future__ import annotations
 
-import os
+import tempfile
+from pathlib import Path
+from typing import TYPE_CHECKING
 
-from ..run import docker_exec
+from baudolo.restore.run import docker_exec
+
+from .version import guard
+
+if TYPE_CHECKING:
+    from collections.abc import Iterable, Iterator
+
+_SUPERUSER_ONLY_PREFIXES = (b"COMMENT ON EXTENSION", b"ALTER DEFAULT PRIVILEGES")
+_EMPTY_PRECLEAN_SQL = Path(__file__).parent / "empty_preclean.sql"
+
+
+def filter_superuser_only_lines(lines: Iterable[bytes]) -> Iterator[bytes]:
+    """Drop superuser-only statements an app-level psql replay cannot run.
+
+    Args:
+        lines: dump lines including their trailing newlines.
+
+    Yields:
+        Every line except top-level statements starting with a superuser-only
+        prefix. Lines inside COPY ... FROM stdin data blocks are passed
+        through untouched: a data row may legally start with the same bytes,
+        and dropping it would silently corrupt the restored table.
+    """
+    in_copy = False
+    for line in lines:
+        if in_copy:
+            yield line
+            if line.rstrip(b"\r\n") == b"\\.":
+                in_copy = False
+            continue
+        if line.startswith(b"COPY ") and line.rstrip(b"\r\n").endswith(b"FROM stdin;"):
+            in_copy = True
+            yield line
+            continue
+        if line.startswith(_SUPERUSER_ONLY_PREFIXES):
+            continue
+        yield line
 
 
 def restore_postgres_sql(
@@ -13,28 +51,25 @@ def restore_postgres_sql(
     password: str,
     sql_path: str,
     empty: bool,
+    check_version: bool = True,
 ) -> None:
-    if not os.path.isfile(sql_path):
+    if not Path(sql_path).is_file():
         raise FileNotFoundError(sql_path)
 
-    # Make password available INSIDE the container for psql.
+    if check_version:
+        guard(
+            sql_path=sql_path,
+            engine="postgres",
+            container=container,
+            user=user,
+            password=password,
+        )
+
     docker_env = {"PGPASSWORD": password}
 
     if empty:
-        drop_sql = r"""
-DO $$ DECLARE r RECORD;
-BEGIN
-  FOR r IN (
-    SELECT table_name AS name, 'TABLE' AS type FROM information_schema.tables WHERE table_schema='public'
-    UNION ALL
-    SELECT routine_name AS name, 'FUNCTION' AS type FROM information_schema.routines WHERE specific_schema='public'
-    UNION ALL
-    SELECT sequence_name AS name, 'SEQUENCE' AS type FROM information_schema.sequences WHERE sequence_schema='public'
-  ) LOOP
-    EXECUTE format('DROP %s public.%I CASCADE', r.type, r.name);
-  END LOOP;
-END $$;
-"""
+        with _EMPTY_PRECLEAN_SQL.open(encoding="utf-8") as preclean:
+            drop_sql = preclean.read()
         docker_exec(
             container,
             ["psql", "-v", "ON_ERROR_STOP=1", "-U", user, "-d", db_name],
@@ -42,11 +77,26 @@ END $$;
             docker_env=docker_env,
         )
 
-    with open(sql_path, "rb") as f:
+    # Filter into a spooled temp file instead of building the whole dump in
+    # memory: production dumps reach many GB and the previous read/splitlines/
+    # join needed roughly three times the dump size in RSS.
+    with Path(sql_path).open("rb") as src, tempfile.TemporaryFile() as filtered:
+        for line in filter_superuser_only_lines(src):
+            filtered.write(line)
+        filtered.seek(0)
         docker_exec(
             container,
-            ["psql", "-v", "ON_ERROR_STOP=1", "-U", user, "-d", db_name],
-            stdin=f,
+            [
+                "psql",
+                "--single-transaction",
+                "-v",
+                "ON_ERROR_STOP=1",
+                "-U",
+                user,
+                "-d",
+                db_name,
+            ],
+            stdin=filtered,
             docker_env=docker_env,
         )
 

src/baudolo/restore/db/version.py

@@ -0,0 +1,147 @@
+"""Refuse a dump the target engine is too old to read.
+
+A restore with ``--empty`` destroys before it replays: the pre-clean drops the
+schema in one session and the dump goes in the next, with no rollback across
+the two. A dump the engine cannot parse therefore does not fail harmlessly -
+it leaves an emptied database behind. Comparing the two versions first turns
+that into a refusal.
+
+Both engines state their origin in the dump's own header, and they do not
+state it the same way. Postgres writes ``-- Dumped from database version``
+around line seven. MariaDB opens line two with ``-- MariaDB dump 10.19-11.8.8``,
+where the first number is mariadb-dump's own version, and names the server only
+further down on the tab-separated ``-- Server version`` line. Matching the first
+number in the header would read the tool on one engine and the server on the
+other, so each engine gets its own pattern.
+
+A ``pg_dumpall`` cluster dump has no version line of its own: its header opens
+with the cluster banner and the roles section, and the first
+``-- Dumped from database version`` belongs to the first database's embedded
+``pg_dump`` output, arbitrarily far down. Hence the scan runs to
+``SCAN_LINES`` rather than to a header-sized handful.
+"""
+
+from __future__ import annotations
+
+import re
+from pathlib import Path
+
+from baudolo.restore.run import docker_exec, stdout_of
+
+SCAN_LINES = 2000
+DUMP_VERSION = {
+    "postgres": re.compile(r"^-- Dumped from database version (\S+)"),
+    "mariadb": re.compile(r"^-- Server version\s+(\S+)"),
+}
+
+
+class VersionMismatchError(Exception):
+    """The dump cannot be replayed into this engine."""
+
+
+def major_of(version: str) -> int:
+    """The major number of an engine version string.
+
+    Args:
+        version: as the engine spells it, e.g. ``17.11`` or
+            ``11.8.8-MariaDB-ubu2404``.
+
+    Raises:
+        VersionMismatchError: the string does not start with a number.
+    """
+    leading = re.match(r"(\d+)", version)
+    if not leading:
+        raise VersionMismatchError(f"cannot read a major version from '{version}'")
+    return int(leading.group(1))
+
+
+def dump_version(sql_path: str, engine: str) -> str:
+    """Read the engine version a dump was taken from, out of its own header.
+
+    Args:
+        sql_path: the dump to read.
+        engine: ``postgres`` or ``mariadb``.
+
+    Returns:
+        The version string as the dump spells it.
+
+    Raises:
+        VersionMismatchError: no version line within the first ``SCAN_LINES``.
+    """
+    pattern = DUMP_VERSION[engine]
+    with Path(sql_path).open(encoding="utf-8", errors="replace") as handle:
+        for _ in range(SCAN_LINES):
+            line = handle.readline()
+            if not line:
+                break
+            found = pattern.search(line)
+            if found:
+                return found.group(1)
+    raise VersionMismatchError(
+        f"{sql_path} carries no {engine} version header in its first {SCAN_LINES} lines"
+    )
+
+
+def server_version(
+    container: str, engine: str, user: str, password: str, client: str = ""
+) -> str:
+    """Ask the running engine which version it is."""
+    if engine == "postgres":
+        return stdout_of(
+            docker_exec(
+                container,
+                ["psql", "-U", user, "-tAc", "SHOW server_version"],
+                capture=True,
+                docker_env={"PGPASSWORD": password},
+            )
+        )
+    return stdout_of(
+        docker_exec(
+            container,
+            [
+                client or "mariadb",
+                "-u",
+                user,
+                f"--password={password}",
+                "-N",
+                "-B",
+                "-e",
+                "SELECT VERSION()",
+            ],
+            capture=True,
+        )
+    )
+
+
+def assert_replayable(sql_path: str, engine: str, dumped: str, serving: str) -> None:
+    """Refuse a dump from a newer major version than the target engine.
+
+    Restoring forward across a major version is the upgrade path and stays
+    allowed; backward is refused, because a newer dump uses syntax an older
+    server rejects and the pre-clean would already have dropped the schema.
+
+    Raises:
+        VersionMismatchError: the dump is newer than the engine.
+    """
+    if major_of(dumped) > major_of(serving):
+        raise VersionMismatchError(
+            f"{sql_path} came from {engine} {dumped} but {serving} is running; "
+            "a newer dump does not replay into an older engine, and --empty "
+            "would drop the schema before finding out"
+        )
+
+
+def guard(
+    *,
+    sql_path: str,
+    engine: str,
+    container: str,
+    user: str,
+    password: str,
+    client: str = "",
+) -> None:
+    """Compare the dump's origin against the running engine before replaying."""
+    dumped = dump_version(sql_path, engine)
+    serving = server_version(container, engine, user, password, client)
+    assert_replayable(sql_path, engine, dumped, serving)
+    print(f"OK: dump is from {engine} {dumped}, {serving} is serving.")

src/baudolo/restore/files.py

@@ -1,13 +1,28 @@
+"""Restore a volume's file tree by writing into its mountpoint.
+
+That shortcut only holds for a plain local volume, where the mountpoint *is*
+the storage. A volume with driver options - NFS, a bind device, tmpfs - keeps
+the same ``/var/lib/docker/volumes/<name>/_data`` path, but docker mounts the
+real backing store over it on demand and unmounts it again when the last
+consumer stops. Writing there while nothing has it mounted lands in the empty
+directory underneath, is hidden by the next mount, and rsync reports success.
+"""
+
 from __future__ import annotations
 
 import os
 import sys
+from pathlib import Path
 
-from .run import run, docker_volume_exists
+from .run import docker_volume_exists, run, stdout_of
+
+INSPECT_FORMAT = (
+    "{{ .Mountpoint }}|{{ .Driver }}|{{ if .Options }}opts{{ else }}plain{{ end }}"
+)
 
 
-def restore_volume_files(volume_name: str, backup_files_dir: str, *, rsync_image: str) -> int:
-    if not os.path.isdir(backup_files_dir):
+def restore_volume_files(volume_name: str, backup_files_dir: str) -> int:
+    if not Path(backup_files_dir).is_dir():
         print(f"ERROR: backup files dir not found: {backup_files_dir}", file=sys.stderr)
         return 2
 
@@ -17,21 +32,33 @@ def restore_volume_files(volume_name: str, backup_files_dir: str, *, rsync_image
     else:
         print(f"Volume {volume_name} already exists.")
 
-    # Keep behavior close to the old script: rsync -avv --delete
-    run(
-        [
-            "docker",
-            "run",
-            "--rm",
-            "-v",
-            f"{volume_name}:/recover/",
-            "-v",
-            f"{backup_files_dir}:/backup/",
-            rsync_image,
-            "sh",
-            "-lc",
-            "rsync -avv --delete /backup/ /recover/",
-        ]
+    cp = run(
+        ["docker", "volume", "inspect", "--format", INSPECT_FORMAT, volume_name],
+        capture=True,
     )
+    fields = stdout_of(cp).split("|")
+    mountpoint = fields[0] if fields else ""
+    if not mountpoint:
+        print(
+            f"ERROR: could not resolve mountpoint for volume {volume_name}",
+            file=sys.stderr,
+        )
+        return 2
+
+    driver, options = ([*fields, "local", "plain"])[1:3]
+    if (driver != "local" or options == "opts") and not os.path.ismount(mountpoint):
+        print(
+            f"ERROR: volume {volume_name} has a backing store of its own "
+            f"(driver {driver}) but nothing has it mounted; writing to "
+            f"{mountpoint} now would land under the mount and be lost. "
+            "Start a container that mounts the volume, then restore again.",
+            file=sys.stderr,
+        )
+        return 2
+
+    # rsync reads "dir/" as its contents and "dir" as the directory itself.
+    src = f"{Path(backup_files_dir)}{os.sep}"
+    dest = f"{Path(mountpoint)}{os.sep}"
+    run(["rsync", "-avv", "--delete", src, dest])
     print("File restore complete.")
     return 0

src/baudolo/restore/paths.py

@@ -1,7 +1,9 @@
 from __future__ import annotations
 
-import os
 from dataclasses import dataclass
+from pathlib import Path
+
+from baudolo.generation import CLUSTER_SUFFIX, DUMP_SUFFIX, FILES_DIR, SQL_DIR
 
 
 @dataclass(frozen=True)
@@ -14,16 +16,20 @@ class BackupPaths:
 
     def root(self) -> str:
         # Always build an absolute path under backups_dir
-        return os.path.join(
-            self.backups_dir,
-            self.backup_hash,
-            self.repo_name,
-            self.version,
-            self.volume_name,
+        return str(
+            Path(self.backups_dir)
+            / self.backup_hash
+            / self.repo_name
+            / self.version
+            / self.volume_name
         )
 
     def files_dir(self) -> str:
-        return os.path.join(self.root(), "files")
+        return str(Path(self.root()) / FILES_DIR)
 
     def sql_file(self, db_name: str) -> str:
-        return os.path.join(self.root(), "sql", f"{db_name}.backup.sql")
+        return str(Path(self.root()) / SQL_DIR / f"{db_name}{DUMP_SUFFIX}")
+
+    def cluster_file(self, instance: str) -> str:
+        """The pg_dumpall stream a `database = '*'` row produces."""
+        return str(Path(self.root()) / SQL_DIR / f"{instance}{CLUSTER_SUFFIX}")

src/baudolo/restore/run.py

@@ -2,7 +2,6 @@ from __future__ import annotations
 
 import subprocess
 import sys
-from typing import Optional
 
 
 def run(
@@ -10,7 +9,7 @@ def run(
     *,
     stdin=None,
     capture: bool = False,
-    env: Optional[dict] = None,
+    env: dict | None = None,
 ) -> subprocess.CompletedProcess:
     try:
         kwargs: dict = {
@@ -26,32 +25,35 @@ def run(
         else:
             kwargs["stdin"] = stdin
 
-        return subprocess.run(cmd, **kwargs)
+        return subprocess.run(cmd, **kwargs)  # noqa: PLW1510 - check lives in kwargs
 
     except subprocess.CalledProcessError as e:
         msg = f"ERROR: command failed ({e.returncode}): {' '.join(cmd)}"
         print(msg, file=sys.stderr)
-        if e.stdout:
+        for stream in (e.stdout, e.stderr):
+            if not stream:
+                continue
             try:
-                print(e.stdout.decode(), file=sys.stderr)
-            except Exception:
-                print(e.stdout, file=sys.stderr)
-        if e.stderr:
-            try:
-                print(e.stderr.decode(), file=sys.stderr)
-            except Exception:
-                print(e.stderr, file=sys.stderr)
+                print(stream.decode(), file=sys.stderr)
+            except (UnicodeDecodeError, AttributeError):
+                print(stream, file=sys.stderr)
         raise
 
 
+def stdout_of(completed: subprocess.CompletedProcess) -> str:
+    """The captured stdout as stripped text, whether it came back bytes or str."""
+    raw = completed.stdout or b""
+    return (raw.decode() if isinstance(raw, bytes) else raw).strip()
+
+
 def docker_exec(
     container: str,
     argv: list[str],
     *,
     stdin=None,
     capture: bool = False,
-    env: Optional[dict] = None,
-    docker_env: Optional[dict[str, str]] = None,
+    env: dict | None = None,
+    docker_env: dict[str, str] | None = None,
 ) -> subprocess.CompletedProcess:
     cmd: list[str] = ["docker", "exec", "-i"]
     if docker_env:
@@ -67,8 +69,8 @@ def docker_exec_sh(
     *,
     stdin=None,
     capture: bool = False,
-    env: Optional[dict] = None,
-    docker_env: Optional[dict[str, str]] = None,
+    env: dict | None = None,
+    docker_env: dict[str, str] | None = None,
 ) -> subprocess.CompletedProcess:
     return docker_exec(
         container,
@@ -85,5 +87,6 @@ def docker_volume_exists(volume: str) -> bool:
         ["docker", "volume", "inspect", volume],
         stdout=subprocess.DEVNULL,
         stderr=subprocess.DEVNULL,
+        check=False,
     )
     return p.returncode == 0

src/baudolo/seed/__main__.py

@@ -1,50 +1,94 @@
-import pandas as pd
+from __future__ import annotations
+
 import argparse
-import os
+import sys
+from pathlib import Path
 
-def check_and_add_entry(file_path, instance, database, username, password):
-    # Check if the file exists and is not empty
-    if os.path.exists(file_path) and os.path.getsize(file_path) > 0:
-        # Read the existing CSV file with header
-        df = pd.read_csv(file_path, sep=';')
+import pandas as pd
+from pandas.errors import EmptyDataError
+
+from baudolo.databases import COLUMNS, DELIMITER, validate_database
+
+
+def _empty_df() -> pd.DataFrame:
+    return pd.DataFrame(columns=list(COLUMNS))
+
+
+def check_and_add_entry(
+    file_path: str,
+    instance: str,
+    database: str | None,
+    username: str,
+    password: str,
+) -> None:
+    """
+    Add or update an entry in databases.csv.
+
+    The function enforces strict validation:
+    - database MUST be set
+    - database MUST be '*' or a valid database name
+    """
+    database = validate_database(database, instance=instance)
+
+    if Path(file_path).exists():
+        try:
+            df = pd.read_csv(
+                file_path,
+                sep=DELIMITER,
+                dtype=str,
+                keep_default_na=False,
+            )
+        except EmptyDataError:
+            print(
+                f"WARNING: databases.csv exists but is empty: {file_path}. Creating header columns.",
+                file=sys.stderr,
+            )
+            df = _empty_df()
     else:
-        # Create a new DataFrame with columns if file does not exist
-        df = pd.DataFrame(columns=['instance', 'database', 'username', 'password'])
+        df = _empty_df()
+    mask = (df["instance"] == instance) & (df["database"] == database)
 
-    # Check if the entry exists and remove it
-    mask = (
-        (df['instance'] == instance) &
-        ((df['database'] == database) |
-        (((df['database'].isna()) | (df['database'] == '')) & (database == ''))) &
-        (df['username'] == username)
-    )
-
-    if not df[mask].empty:
-        print("Replacing existing entry.")
-        df = df[~mask]
+    if mask.any():
+        print("Updating existing entry.")
+        df.loc[mask, ["username", "password"]] = [username, password]
     else:
         print("Adding new entry.")
+        new_entry = pd.DataFrame(
+            [[instance, database, username, password]],
+            columns=list(COLUMNS),
+        )
+        df = pd.concat([df, new_entry], ignore_index=True)
 
-    # Create a new DataFrame for the new entry
-    new_entry = pd.DataFrame([{'instance': instance, 'database': database, 'username': username, 'password': password}])
+    df.to_csv(file_path, sep=DELIMITER, index=False)
 
-    # Add (or replace) the entry using concat
-    df = pd.concat([df, new_entry], ignore_index=True)
 
-    # Save the updated CSV file
-    df.to_csv(file_path, sep=';', index=False)
-
-def main():
-    parser = argparse.ArgumentParser(description="Check and replace (or add) a database entry in a CSV file.")
-    parser.add_argument("file_path", help="Path to the CSV file")
-    parser.add_argument("instance", help="Database instance")
-    parser.add_argument("database", help="Database name")
-    parser.add_argument("username", help="Username")
-    parser.add_argument("password", nargs='?', default="", help="Password (optional)")
+def main() -> None:
+    parser = argparse.ArgumentParser(
+        description="Seed or update databases.csv for backup configuration."
+    )
+    parser.add_argument("file", help="Path to databases.csv")
+    parser.add_argument("instance", help="Instance name (e.g. bigbluebutton)")
+    parser.add_argument(
+        "database",
+        help="Database name or '*' to dump all databases",
+    )
+    parser.add_argument("username", help="Database username")
+    parser.add_argument("password", help="Database password")
 
     args = parser.parse_args()
 
-    check_and_add_entry(args.file_path, args.instance, args.database, args.username, args.password)
+    try:
+        check_and_add_entry(
+            file_path=args.file,
+            instance=args.instance,
+            database=args.database,
+            username=args.username,
+            password=args.password,
+        )
+    except Exception as exc:  # noqa: BLE001 - CLI boundary: any failure becomes exit 1
+        print(f"ERROR: {exc}", file=sys.stderr)
+        sys.exit(1)
+
 
 if __name__ == "__main__":
     main()

tests/e2e/faithful_driver.py

@@ -0,0 +1,90 @@
+"""Show what a real snapshot holds for a volume that has its own storage.
+
+Runs inside the privileged container that built the btrfs subject. Prints one
+PASS/FAIL line per assertion and exits non-zero on the first failure.
+"""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+from pathlib import Path
+
+sys.path.insert(0, "/src")
+
+from baudolo.backup.snapshot import (
+    SnapshotError,
+    snapshot_source,
+    unsnapshotted,
+    volume_snapshot,
+)
+from baudolo.backup.volume import Backing
+
+SUBJECT = sys.argv[1]
+
+
+def shell(command: list[str]) -> list[str]:
+    proc = subprocess.run(command, capture_output=True, text=True, check=False)
+    if proc.returncode != 0:
+        raise SnapshotError(
+            f"{' '.join(command)} exited {proc.returncode}: {proc.stderr.strip()}"
+        )
+    return proc.stdout.splitlines()
+
+
+def check(label: str, condition: bool) -> None:
+    print(f"{'PASS' if condition else 'FAIL'} {label}", flush=True)
+    if not condition:
+        sys.exit(1)
+
+
+def volume(name: str, payload: str) -> Path:
+    path = Path(SUBJECT) / "volumes" / name / "_data"
+    path.mkdir(parents=True, exist_ok=True)
+    (path / "state").write_text(payload)
+    return path
+
+
+plain = volume("plain", "plain-payload")
+own = Path(SUBJECT) / "volumes" / "own" / "_data"
+own.mkdir(parents=True, exist_ok=True)
+shell(["mount", "-t", "tmpfs", "tmpfs", own])
+(own / "state").write_text("own-payload")
+
+check("a plain volume is captured", unsnapshotted(Backing(str(plain)), SUBJECT) is None)
+check(
+    "a volume on a mount of its own is not",
+    unsnapshotted(Backing(str(own)), SUBJECT) is not None,
+)
+check(
+    "a declared backing store is not, mounted or not",
+    unsnapshotted(Backing(str(plain), options={"type": "nfs"}), SUBJECT) is not None,
+)
+check(
+    "a foreign driver is not",
+    unsnapshotted(Backing(str(plain), driver="rexray"), SUBJECT) is not None,
+)
+
+with volume_snapshot("btrfs", SUBJECT, "e2e", run=shell) as resolve:
+    frozen_plain = Path(resolve(str(plain)))
+    frozen_own = Path(resolve(str(own)))
+
+    check(
+        "the snapshot carries the plain volume",
+        (frozen_plain / "state").read_text() == "plain-payload",
+    )
+    check(
+        "the snapshot shows the other volume as an empty directory",
+        frozen_own.is_dir() and not any(frozen_own.iterdir()),
+    )
+
+    source, reason = snapshot_source(resolve, Backing(str(plain)), SUBJECT)
+    check(
+        "the plain volume is read from the snapshot",
+        source is not None and source.rstrip("/") == str(frozen_plain),
+    )
+
+    source, reason = snapshot_source(resolve, Backing(str(own)), SUBJECT)
+    check(f"the other volume degrades to live: {reason[:60]}", source is None)
+
+print("ALL OK", flush=True)

tests/e2e/helpers.py

@@ -1,222 +0,0 @@
-# tests/e2e/helpers.py
-from __future__ import annotations
-
-import shutil
-import subprocess
-import time
-import uuid
-from pathlib import Path
-
-
-def run(
-    cmd: list[str],
-    *,
-    capture: bool = True,
-    check: bool = True,
-    cwd: str | None = None,
-) -> subprocess.CompletedProcess:
-    try:
-        return subprocess.run(
-            cmd,
-            check=check,
-            cwd=cwd,
-            text=True,
-            capture_output=capture,
-        )
-    except subprocess.CalledProcessError as e:
-        # Print captured output so failing E2E tests are "live" / debuggable in CI logs
-        print(">>> command failed:", " ".join(cmd))
-        print(">>> exit code:", e.returncode)
-        if e.stdout:
-            print(">>> STDOUT:\n" + e.stdout)
-        if e.stderr:
-            print(">>> STDERR:\n" + e.stderr)
-        raise
-
-
-def sh(cmd: str, *, capture: bool = True, check: bool = True) -> subprocess.CompletedProcess:
-    return run(["sh", "-lc", cmd], capture=capture, check=check)
-
-
-def unique(prefix: str) -> str:
-    return f"{prefix}-{uuid.uuid4().hex[:10]}"
-
-
-def require_docker() -> None:
-    run(["docker", "version"], capture=True, check=True)
-
-
-def machine_hash() -> str:
-    out = sh("sha256sum /etc/machine-id | awk '{print $1}'").stdout.strip()
-    if len(out) < 16:
-        raise RuntimeError("Could not determine machine hash from /etc/machine-id")
-    return out
-
-
-def wait_for_log(container: str, pattern: str, timeout_s: int = 60) -> None:
-    deadline = time.time() + timeout_s
-    while time.time() < deadline:
-        p = run(["docker", "logs", container], capture=True, check=False)
-        if pattern in (p.stdout or ""):
-            return
-        time.sleep(1)
-    raise TimeoutError(f"Timed out waiting for log pattern '{pattern}' in {container}")
-
-
-def wait_for_postgres(container: str, *, user: str = "postgres", timeout_s: int = 90) -> None:
-    """
-    Docker-outside-of-Docker friendly readiness: check from inside the DB container.
-    """
-    deadline = time.time() + timeout_s
-    while time.time() < deadline:
-        p = run(
-            ["docker", "exec", container, "sh", "-lc", f"pg_isready -U {user} -h localhost"],
-            capture=True,
-            check=False,
-        )
-        if p.returncode == 0:
-            return
-        time.sleep(1)
-    raise TimeoutError(f"Timed out waiting for Postgres readiness in container {container}")
-
-
-def wait_for_mariadb(container: str, *, root_password: str, timeout_s: int = 90) -> None:
-    """
-    Liveness probe for MariaDB.
-
-    IMPORTANT (MariaDB 11):
-    Root TCP auth is often restricted (unix_socket auth), so a TCP ping like
-    `mariadb-admin -uroot -p... -h localhost ping` can fail even though the server is up.
-    We therefore check readiness via a socket-based query.
-    """
-    deadline = time.time() + timeout_s
-    while time.time() < deadline:
-        p = run(
-            ["docker", "exec", container, "sh", "-lc", "mariadb -uroot --protocol=socket -e \"SELECT 1;\""],
-            capture=True,
-            check=False,
-        )
-        if p.returncode == 0:
-            return
-        time.sleep(1)
-    raise TimeoutError(f"Timed out waiting for MariaDB readiness in container {container}")
-
-
-def wait_for_mariadb_sql(container: str, *, user: str, password: str, timeout_s: int = 90) -> None:
-    """
-    SQL login readiness for the *dedicated test user* over TCP.
-
-    This is separate from wait_for_mariadb(root) because root may be socket-only,
-    while the tests use a normal user that should work via TCP.
-    """
-    deadline = time.time() + timeout_s
-    while time.time() < deadline:
-        p = run(
-            [
-                "docker",
-                "exec",
-                container,
-                "sh",
-                "-lc",
-                f"mariadb -h 127.0.0.1 -u{user} -p{password} -e \"SELECT 1;\"",
-            ],
-            capture=True,
-            check=False,
-        )
-        if p.returncode == 0:
-            return
-        time.sleep(1)
-    raise TimeoutError(f"Timed out waiting for MariaDB SQL login readiness in container {container}")
-
-
-def backup_run(
-    *,
-    backups_dir: str,
-    repo_name: str,
-    compose_dir: str,
-    databases_csv: str,
-    database_containers: list[str],
-    images_no_stop_required: list[str],
-    images_no_backup_required: list[str] | None = None,
-    dump_only: bool = False,
-) -> None:
-    cmd = [
-        "baudolo",
-        "--compose-dir", compose_dir,
-        "--docker-compose-hard-restart-required", "mailu",
-        "--repo-name", repo_name,
-        "--databases-csv", databases_csv,
-        "--backups-dir", backups_dir,
-        "--database-containers", *database_containers,
-        "--images-no-stop-required", *images_no_stop_required,
-    ]
-    if images_no_backup_required:
-        cmd += ["--images-no-backup-required", *images_no_backup_required]
-    if dump_only:
-        cmd += ["--dump-only"]
-
-    try:
-        run(cmd, capture=True, check=True)
-    except subprocess.CalledProcessError as e:
-        print(">>> baudolo failed (exit code:", e.returncode, ")")
-        if e.stdout:
-            print(">>> baudolo STDOUT:\n" + e.stdout)
-        if e.stderr:
-            print(">>> baudolo STDERR:\n" + e.stderr)
-        raise
-
-
-def latest_version_dir(backups_dir: str, repo_name: str) -> tuple[str, str]:
-    """
-    Returns (hash, version) for the latest backup.
-    """
-    h = machine_hash()
-    root = Path(backups_dir) / h / repo_name
-    if not root.is_dir():
-        raise FileNotFoundError(str(root))
-
-    versions = sorted([p.name for p in root.iterdir() if p.is_dir()])
-    if not versions:
-        raise RuntimeError(f"No versions found under {root}")
-    return h, versions[-1]
-
-
-def backup_path(backups_dir: str, repo_name: str, version: str, volume: str) -> Path:
-    h = machine_hash()
-    return Path(backups_dir) / h / repo_name / version / volume
-
-
-def create_minimal_compose_dir(base: str) -> str:
-    """
-    baudolo requires --compose-dir. Create an empty dir with one non-compose subdir.
-    """
-    p = Path(base) / "compose-root"
-    p.mkdir(parents=True, exist_ok=True)
-    (p / "noop").mkdir(parents=True, exist_ok=True)
-    return str(p)
-
-
-def write_databases_csv(path: str, rows: list[tuple[str, str, str, str]]) -> None:
-    """
-    rows: (instance, database, username, password)
-    database may be '' (empty) to trigger pg_dumpall behavior if you want, but here we use db name.
-    """
-    Path(path).parent.mkdir(parents=True, exist_ok=True)
-    with open(path, "w", encoding="utf-8") as f:
-        f.write("instance;database;username;password\n")
-        for inst, db, user, pw in rows:
-            f.write(f"{inst};{db};{user};{pw}\n")
-
-
-def cleanup_docker(*, containers: list[str], volumes: list[str]) -> None:
-    for c in containers:
-        run(["docker", "rm", "-f", c], capture=True, check=False)
-    for v in volumes:
-        run(["docker", "volume", "rm", "-f", v], capture=True, check=False)
-
-
-def ensure_empty_dir(path: str) -> None:
-    p = Path(path)
-    if p.exists():
-        shutil.rmtree(p)
-    p.mkdir(parents=True, exist_ok=True)

tests/e2e/helpers/__init__.py

@@ -0,0 +1,4 @@
+"""Shared e2e helpers, re-exported so tests import one name."""
+
+from .fixtures import *
+from .process import *

tests/e2e/helpers/fixtures.py

@@ -0,0 +1,114 @@
+"""Fixtures and paths the e2e suite builds its scenarios from."""
+
+from __future__ import annotations
+
+import shutil
+import subprocess
+from pathlib import Path
+
+from .process import machine_hash, run
+
+# postgres 18+ mounts at /var/lib/postgresql, not /var/lib/postgresql/data.
+POSTGRES_IMAGE = "postgres:alpine"
+POSTGRES_DATA_DIR = "/var/lib/postgresql"
+MARIADB_IMAGE = "mariadb:latest"
+MARIADB_DATA_DIR = "/var/lib/mysql"
+
+
+def backup_run(
+    *,
+    backups_dir: str,
+    repo_name: str,
+    compose_dir: str,
+    databases_csv: str,
+    database_containers: list[str],
+    images_no_stop_required: list[str],
+    images_no_backup_required: list[str] | None = None,
+    only_sql: bool = False,
+) -> None:
+    cmd = [
+        "baudolo",
+        "--compose-dir",
+        compose_dir,
+        "--hard-restart-projects",
+        "mailu",
+        "--repo-name",
+        repo_name,
+        "--databases-csv",
+        databases_csv,
+        "--backups-dir",
+        backups_dir,
+        "--database-containers",
+        *database_containers,
+        "--images-no-stop-required",
+        *images_no_stop_required,
+    ]
+    if images_no_backup_required:
+        cmd += ["--images-no-backup-required", *images_no_backup_required]
+    if only_sql:
+        cmd += ["--only-sql"]
+
+    try:
+        run(cmd, capture=True, check=True)
+    except subprocess.CalledProcessError as e:
+        print(">>> baudolo failed (exit code:", e.returncode, ")")
+        if e.stdout:
+            print(">>> baudolo STDOUT:\n" + e.stdout)
+        if e.stderr:
+            print(">>> baudolo STDERR:\n" + e.stderr)
+        raise
+
+
+def latest_version_dir(backups_dir: str, repo_name: str) -> tuple[str, str]:
+    """
+    Returns (hash, version) for the latest backup.
+    """
+    h = machine_hash()
+    root = Path(backups_dir) / h / repo_name
+    if not root.is_dir():
+        raise FileNotFoundError(str(root))
+
+    versions = sorted([p.name for p in root.iterdir() if p.is_dir()])
+    if not versions:
+        raise RuntimeError(f"No versions found under {root}")
+    return h, versions[-1]
+
+
+def backup_path(backups_dir: str, repo_name: str, version: str, volume: str) -> Path:
+    h = machine_hash()
+    return Path(backups_dir) / h / repo_name / version / volume
+
+
+def create_minimal_compose_dir(base: str) -> str:
+    """
+    baudolo requires --compose-dir. Create an empty dir with one non-compose subdir.
+    """
+    p = Path(base) / "compose-root"
+    p.mkdir(parents=True, exist_ok=True)
+    (p / "noop").mkdir(parents=True, exist_ok=True)
+    return str(p)
+
+
+def write_databases_csv(path: str, rows: list[tuple[str, str, str, str]]) -> None:
+    """
+    rows: (instance, database, username, password)
+    database may be '' (empty) to trigger pg_dumpall behavior if you want, but here we use db name.
+    """
+    Path(path).parent.mkdir(parents=True, exist_ok=True)
+    with Path(path).open("w", encoding="utf-8") as f:
+        f.write("instance;database;username;password\n")
+        f.writelines(f"{inst};{db};{user};{pw}\n" for inst, db, user, pw in rows)
+
+
+def cleanup_docker(*, containers: list[str], volumes: list[str]) -> None:
+    for c in containers:
+        run(["docker", "rm", "-f", c], capture=True, check=False)
+    for v in volumes:
+        run(["docker", "volume", "rm", "-f", v], capture=True, check=False)
+
+
+def ensure_empty_dir(path: str) -> None:
+    p = Path(path)
+    if p.exists():
+        shutil.rmtree(p)
+    p.mkdir(parents=True, exist_ok=True)

tests/e2e/helpers/process.py

@@ -0,0 +1,156 @@
+"""Process, docker and readiness helpers for the e2e suite."""
+
+from __future__ import annotations
+
+import subprocess
+import time
+import uuid
+
+
+def run(
+    cmd: list[str],
+    *,
+    capture: bool = True,
+    check: bool = True,
+    cwd: str | None = None,
+) -> subprocess.CompletedProcess:
+    try:
+        return subprocess.run(
+            cmd,
+            check=check,
+            cwd=cwd,
+            text=True,
+            capture_output=capture,
+        )
+    except subprocess.CalledProcessError as e:
+        # Print captured output so failing E2E tests are "live" / debuggable in CI logs
+        print(">>> command failed:", " ".join(cmd))
+        print(">>> exit code:", e.returncode)
+        if e.stdout:
+            print(">>> STDOUT:\n" + e.stdout)
+        if e.stderr:
+            print(">>> STDERR:\n" + e.stderr)
+        raise
+
+
+def sh(
+    cmd: str, *, capture: bool = True, check: bool = True
+) -> subprocess.CompletedProcess:
+    return run(["sh", "-lc", cmd], capture=capture, check=check)
+
+
+def unique(prefix: str) -> str:
+    return f"{prefix}-{uuid.uuid4().hex[:10]}"
+
+
+def require_docker() -> None:
+    run(["docker", "version"], capture=True, check=True)
+
+
+def machine_hash() -> str:
+    out = sh("sha256sum /etc/machine-id | awk '{print $1}'").stdout.strip()
+    if len(out) < 16:
+        raise RuntimeError("Could not determine machine hash from /etc/machine-id")
+    return out
+
+
+def wait_for_log(container: str, pattern: str, timeout_s: int = 60) -> None:
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        p = run(["docker", "logs", container], capture=True, check=False)
+        if pattern in (p.stdout or ""):
+            return
+        time.sleep(1)
+    raise TimeoutError(f"Timed out waiting for log pattern '{pattern}' in {container}")
+
+
+def wait_for_postgres(
+    container: str, *, user: str = "postgres", timeout_s: int = 90
+) -> None:
+    """
+    Docker-outside-of-Docker friendly readiness: check from inside the DB container.
+    """
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        p = run(
+            [
+                "docker",
+                "exec",
+                container,
+                "sh",
+                "-lc",
+                f"pg_isready -U {user} -h localhost",
+            ],
+            capture=True,
+            check=False,
+        )
+        if p.returncode == 0:
+            return
+        time.sleep(1)
+    raise TimeoutError(
+        f"Timed out waiting for Postgres readiness in container {container}"
+    )
+
+
+def wait_for_mariadb(
+    container: str, *, root_password: str, timeout_s: int = 90
+) -> None:
+    """
+    Liveness probe for MariaDB.
+
+    IMPORTANT (MariaDB 11):
+    Root TCP auth is often restricted (unix_socket auth), so a TCP ping like
+    `mariadb-admin -uroot -p... -h localhost ping` can fail even though the server is up.
+    We therefore check readiness via a socket-based query.
+    """
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        p = run(
+            [
+                "docker",
+                "exec",
+                container,
+                "sh",
+                "-lc",
+                'mariadb -uroot --protocol=socket -e "SELECT 1;"',
+            ],
+            capture=True,
+            check=False,
+        )
+        if p.returncode == 0:
+            return
+        time.sleep(1)
+    raise TimeoutError(
+        f"Timed out waiting for MariaDB readiness in container {container}"
+    )
+
+
+def wait_for_mariadb_sql(
+    container: str, *, user: str, password: str, timeout_s: int = 90
+) -> None:
+    """
+    SQL login readiness for the *dedicated test user* over TCP.
+
+    This is separate from wait_for_mariadb(root) because root may be socket-only,
+    while the tests use a normal user that should work via TCP.
+    """
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        p = run(
+            [
+                "docker",
+                "exec",
+                container,
+                "sh",
+                "-lc",
+                f'mariadb -h 127.0.0.1 -u{user} -p{password} -e "SELECT 1;"',
+            ],
+            capture=True,
+            check=False,
+        )
+        if p.returncode == 0:
+            return
+        time.sleep(1)
+    raise TimeoutError(
+        f"Timed out waiting for MariaDB SQL login readiness in container {container}"
+    )

tests/e2e/snapshot_db_driver.py

@@ -0,0 +1,43 @@
+"""Copy a live database's volume out of a snapshot, using the real backup path.
+
+Runs inside the privileged container built by test_e2e_snapshot_db.py, where a
+database is mid-write on a btrfs subvolume. Exercises volume_snapshot and
+backup_volume exactly as a backup run would.
+"""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+
+sys.path.insert(0, "/src")
+
+from baudolo.backup.snapshot import SnapshotError, volume_snapshot
+from baudolo.backup.volume import backup_volume
+
+SUBJECT = "/subject/docker"
+VOLUME = "mariadb_data"
+DATADIR = f"{SUBJECT}/volumes/{VOLUME}/_data"
+VERSIONS = "/backups"
+GENERATION = f"{VERSIONS}/20260731"
+
+
+def shell(command: list[str]) -> list[str]:
+    proc = subprocess.run(command, capture_output=True, text=True, check=False)
+    if proc.returncode != 0:
+        raise SnapshotError(
+            f"{' '.join(command)} exited {proc.returncode}: {proc.stderr.strip()}"
+        )
+    return proc.stdout.splitlines()
+
+
+with volume_snapshot("btrfs", SUBJECT, "dbtest", run=shell) as resolve:
+    backup_volume(
+        VERSIONS,
+        VOLUME,
+        f"{GENERATION}/{VOLUME}",
+        authoritative=True,
+        source=resolve(f"{DATADIR}/"),
+    )
+
+print("SNAPSHOT COPY DONE", flush=True)

tests/e2e/snapshot_driver.py

@@ -0,0 +1,64 @@
+"""Exercise volume_snapshot against a real filesystem, from inside a container.
+
+Runs where loop devices exist. Prints one PASS/FAIL line per assertion and exits
+non-zero on the first failure, so the calling test can surface the reason.
+"""
+
+from __future__ import annotations
+
+import subprocess
+import sys
+from pathlib import Path
+
+sys.path.insert(0, "/src")
+
+from baudolo.backup.snapshot import SnapshotError, volume_snapshot
+
+KIND = sys.argv[1]
+SUBJECT = sys.argv[2]
+EXPECT = sys.argv[3]
+
+
+def shell(command: list[str]) -> list[str]:
+    proc = subprocess.run(command, capture_output=True, text=True, check=False)
+    if proc.returncode != 0:
+        raise SnapshotError(
+            f"{' '.join(command)} exited {proc.returncode}: {proc.stderr.strip()}"
+        )
+    return proc.stdout.splitlines()
+
+
+def check(label: str, condition: bool) -> None:
+    print(f"{'PASS' if condition else 'FAIL'} {label}", flush=True)
+    if not condition:
+        sys.exit(1)
+
+
+volume = Path(SUBJECT) / "volumes" / "demo" / "_data"
+volume.mkdir(parents=True, exist_ok=True)
+(volume / "state").write_text("before\n")
+
+if EXPECT == "unsupported":
+    try:
+        with volume_snapshot(KIND, SUBJECT, "e2e", run=shell):
+            check("snapshot on an unsupported filesystem must not succeed", False)
+    except SnapshotError as exc:
+        check(f"refused loudly: {str(exc)[:60]}", True)
+    sys.exit(0)
+
+with volume_snapshot(KIND, SUBJECT, "e2e", run=shell) as resolve:
+    frozen = Path(resolve(str(volume))) / "state"
+    check("the snapshot exposes the volume", frozen.is_file())
+    check("the snapshot carries the content", frozen.read_text() == "before\n")
+
+    (volume / "state").write_text("after\n")
+    check("a later write does not reach the snapshot", frozen.read_text() == "before\n")
+    check("the live tree did change", (volume / "state").read_text() == "after\n")
+
+    root = Path(resolve(SUBJECT))
+
+check(
+    "the snapshot is removed afterwards",
+    not root.exists() or not (root / "volumes").exists(),
+)
+print("ALL OK", flush=True)

tests/e2e/test_e2e_app_container_ships_client_tools.py

@@ -0,0 +1,163 @@
+"""An application container that ships the engine's client tools.
+
+This is the shape a dedicated database deploys in: the engine runs as
+`<app>-database` while the application itself runs as `<app>`, and neither is
+declared through --database-containers, so both names go through the instance
+regex. `<app>-database` loses its suffix and lands on the instance `<app>` -
+and `<app>` carries no database token at all, so a fallback that returns the
+name unchanged lands on that same instance and offers the application container
+as a second engine for the same row.
+
+Discourse is the live example: its application container is named `discourse`
+by its own launcher and ships pg_dumpall, so a dump command starts there and
+writes a file that looks like a backup and holds none of the data.
+"""
+
+import unittest
+from pathlib import Path
+
+from baudolo.generation import DUMP_SUFFIX, FILES_DIR, SQL_DIR
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+MARKER = "the-application-volume-holds-files"
+PAYLOAD = "shop-payload"
+
+
+class TestE2EAppContainerShipsClientTools(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        # uuid4 hex may begin with "db", which the instance regex would split
+        # on and turn the application container into a different instance,
+        # hiding exactly the collision this module is about.
+        cls.prefix = unique("baudolo-e2e-app-tools").replace("-db", "-xb")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.engine = f"{cls.prefix}-shop-database"
+        cls.app = f"{cls.prefix}-shop"
+        cls.engine_volume = f"{cls.prefix}-shop-database-vol"
+        cls.app_volume = f"{cls.prefix}-shop-app-vol"
+        cls.containers = [cls.engine, cls.app]
+        cls.volumes = [cls.engine_volume, cls.app_volume]
+
+        run(["docker", "volume", "create", cls.engine_volume])
+        run(["docker", "volume", "create", cls.app_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.engine,
+                "-e",
+                "POSTGRES_PASSWORD=shoppw",
+                "-e",
+                "POSTGRES_DB=shopdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.engine_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.app,
+                "--entrypoint",
+                "sh",
+                "-v",
+                f"{cls.app_volume}:/data",
+                POSTGRES_IMAGE,
+                "-c",
+                f"echo '{MARKER}' > /data/marker.txt && sleep 3600",
+            ]
+        )
+
+        wait_for_postgres(cls.engine, user="postgres", timeout_s=90)
+        run(
+            [
+                "docker",
+                "exec",
+                cls.engine,
+                "sh",
+                "-lc",
+                (
+                    'psql -U postgres -d shopdb -c "CREATE TABLE orders (id int, '
+                    f"note text); INSERT INTO orders VALUES (1,'{PAYLOAD}');\""
+                ),
+            ],
+            check=True,
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv,
+            [(cls.app, "shopdb", "postgres", "shoppw")],
+        )
+
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=["dummy-db"],
+            images_no_stop_required=[POSTGRES_IMAGE],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def volume_dir(self, volume: str) -> Path:
+        return backup_path(self.backups_dir, self.repo_name, self.version, volume)
+
+    def test_the_engine_volume_was_dumped(self) -> None:
+        dump = self.volume_dir(self.engine_volume) / SQL_DIR / f"shopdb{DUMP_SUFFIX}"
+        self.assertTrue(dump.is_file(), f"expected a dump at {dump}")
+        self.assertIn(PAYLOAD, dump.read_text(encoding="utf-8"))
+
+    def test_the_application_volume_produced_no_dump(self) -> None:
+        """The collision this module exists for: the application container
+        answers the same instance as the engine and starts a dump of its own."""
+        sql_dir = self.volume_dir(self.app_volume) / SQL_DIR
+        self.assertFalse(
+            sql_dir.exists(),
+            f"the application container was dumped: {sorted(sql_dir.iterdir())}"
+            if sql_dir.exists()
+            else "",
+        )
+
+    def test_the_application_volume_was_backed_up_as_files(self) -> None:
+        """Refusing the dump must not cost the volume its backup."""
+        marker = self.volume_dir(self.app_volume) / FILES_DIR / "marker.txt"
+        self.assertTrue(marker.is_file(), f"expected a file backup at {marker}")
+        self.assertIn(MARKER, marker.read_text(encoding="utf-8"))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_cli_contract_only_sql.py

@@ -0,0 +1,33 @@
+import unittest
+
+from .helpers import run
+
+WITHDRAWN_FLAGS = ["--dump-only", "--dump-only-sql", "--everything"]
+
+
+class TestE2ECLIContractOnlySql(unittest.TestCase):
+    def test_help_mentions_the_flag(self) -> None:
+        cp = run(["baudolo", "--help"], capture=True, check=True)
+        out = (cp.stdout or "") + "\n" + (cp.stderr or "")
+        self.assertIn(
+            "--only-sql",
+            out,
+            f"Expected '--only-sql' to appear in --help output. Output:\n{out}",
+        )
+
+    def test_a_withdrawn_flag_is_rejected(self) -> None:
+        for flag in WITHDRAWN_FLAGS:
+            with self.subTest(flag=flag):
+                cp = run(["baudolo", flag], capture=True, check=False)
+                self.assertEqual(
+                    cp.returncode,
+                    2,
+                    f"Expected exitcode 2 for unknown args, got {cp.returncode}\n"
+                    f"STDOUT={cp.stdout}\nSTDERR={cp.stderr}",
+                )
+                err = (cp.stderr or "") + "\n" + (cp.stdout or "")
+                # Argparse typically prints "unrecognized arguments"
+                self.assertTrue(
+                    ("unrecognized arguments" in err) or ("usage:" in err.lower()),
+                    f"Expected argparse-style error output. Output:\n{err}",
+                )

tests/e2e/test_e2e_engine_detection_by_tool.py

@@ -0,0 +1,177 @@
+"""The engine comes from the tools a container ships, not from its image name.
+
+Two containers in one backup run, each lying in one direction:
+
+* a real Postgres tagged `<prefix>-database`, the way a dedicated database is
+  built inside an app's own stack - no engine token anywhere in the name;
+* an Alpine tagged `postgres:<prefix>`, carrying the token without shipping a
+  single Postgres binary.
+
+Reading the name gets both wrong, and the second one fatally: pg_dump exits 127
+inside Alpine and takes the whole run with it.
+"""
+
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+IMPOSTOR_BASE_IMAGE = "alpine:3.20"
+MARKER = "engine-detection-by-tool"
+
+
+class TestE2EEngineDetectionByTool(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-engine-by-tool")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.engine_image = f"{cls.prefix}-database:17"
+        cls.impostor_image = f"postgres:{cls.prefix}"
+
+        cls.engine_container = f"{cls.prefix}-engine"
+        cls.impostor_container = f"{cls.prefix}-impostor"
+        cls.engine_volume = f"{cls.prefix}-engine-vol"
+        cls.impostor_volume = f"{cls.prefix}-impostor-vol"
+
+        cls.containers = [cls.engine_container, cls.impostor_container]
+        cls.volumes = [cls.engine_volume, cls.impostor_volume]
+
+        run(["docker", "pull", POSTGRES_IMAGE])
+        run(["docker", "pull", IMPOSTOR_BASE_IMAGE])
+        run(["docker", "tag", POSTGRES_IMAGE, cls.engine_image])
+        run(["docker", "tag", IMPOSTOR_BASE_IMAGE, cls.impostor_image])
+        run(["docker", "volume", "create", cls.engine_volume])
+        run(["docker", "volume", "create", cls.impostor_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.engine_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.engine_volume}:{POSTGRES_DATA_DIR}",
+                cls.engine_image,
+            ]
+        )
+        wait_for_postgres(cls.engine_container, user="postgres", timeout_s=90)
+        run(
+            [
+                "docker",
+                "exec",
+                cls.engine_container,
+                "sh",
+                "-lc",
+                (
+                    "psql -U postgres -d appdb -c "
+                    '"CREATE TABLE t (id int primary key, v text); '
+                    "INSERT INTO t VALUES (1,'ok');\""
+                ),
+            ]
+        )
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.impostor_container,
+                "-v",
+                f"{cls.impostor_volume}:/data",
+                cls.impostor_image,
+                "sh",
+                "-lc",
+                f"echo '{MARKER}' > /data/marker.txt && sleep 3600",
+            ]
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv,
+            [
+                (cls.engine_container, "appdb", "postgres", "pgpw"),
+                (cls.impostor_container, "appdb", "postgres", "pgpw"),
+            ],
+        )
+
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.engine_container, cls.impostor_container],
+            images_no_stop_required=[cls.engine_image, cls.impostor_image],
+            only_sql=True,
+        )
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+        run(["docker", "rmi", cls.engine_image], check=False)
+        run(["docker", "rmi", cls.impostor_image], check=False)
+
+    def _volume_dir(self, volume: str):
+        return backup_path(self.backups_dir, self.repo_name, self.version, volume)
+
+    def test_an_engine_without_an_engine_name_is_still_dumped(self) -> None:
+        dump = self._volume_dir(self.engine_volume) / "sql" / "appdb.backup.sql"
+        self.assertTrue(
+            dump.is_file(),
+            f"a Postgres tagged '{self.engine_image}' produced no dump at {dump}",
+        )
+        self.assertIn("Dumped by pg_dump", dump.read_text(encoding="utf-8"))
+
+    def test_an_engine_name_without_an_engine_is_not_dumped(self) -> None:
+        sql_dir = self._volume_dir(self.impostor_volume) / "sql"
+        dumps = list(sql_dir.glob("*.sql")) if sql_dir.exists() else []
+        self.assertEqual(
+            dumps,
+            [],
+            f"'{self.impostor_image}' ships no Postgres yet was dumped: {dumps}",
+        )
+
+    def test_the_recognised_engine_is_dumped_instead_of_copied(self) -> None:
+        files = self._volume_dir(self.engine_volume) / "files"
+        self.assertFalse(
+            files.exists(),
+            f"--only-sql still copied the engine's files to {files}",
+        )
+
+    def test_the_impostor_falls_through_to_a_file_backup(self) -> None:
+        files = self._volume_dir(self.impostor_volume) / "files"
+        self.assertTrue(files.is_dir(), f"expected a file backup at {files}")
+        self.assertEqual(
+            (files / "marker.txt").read_text(encoding="utf-8").strip(), MARKER
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_engine_detection_registry_prefix.py

@@ -0,0 +1,105 @@
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+REGISTRY_HOST = "svc-db-mariadb-swarm-mgr-01:5000"
+
+
+class TestE2EEngineDetectionRegistryPrefix(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-registry-prefix")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.image = f"{REGISTRY_HOST}/postgres_custom:17-3.5"
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "pull", POSTGRES_IMAGE])
+        run(["docker", "tag", POSTGRES_IMAGE, cls.image])
+        run(["docker", "volume", "create", cls.pg_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                cls.image,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                "psql -U postgres -d appdb -c \"CREATE TABLE t (id int primary key, v text); INSERT INTO t VALUES (1,'ok');\"",
+            ]
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")]
+        )
+
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.pg_container],
+            images_no_stop_required=[cls.image],
+        )
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+        run(["docker", "rmi", cls.image], check=False)
+
+    def test_the_registry_host_does_not_pick_the_engine(self) -> None:
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.pg_volume)
+            / "sql"
+            / "appdb.backup.sql"
+        )
+        self.assertTrue(p.is_file(), f"Expected a pg_dump at: {p}")
+        self.assertIn("Dumped by pg_dump", p.read_text(encoding="utf-8"))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_files_full.py

@@ -1,17 +1,16 @@
 import unittest
-from pathlib import Path
 
 from .helpers import (
-    backup_run,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
+    run,
     unique,
     write_databases_csv,
-    run,
 )
 
 
@@ -31,27 +30,32 @@ class TestE2EFilesFull(unittest.TestCase):
         cls.containers = []
         cls.volumes = [cls.volume_src, cls.volume_dst]
 
-        # create source volume with a file
         run(["docker", "volume", "create", cls.volume_src])
-        run([
-            "docker", "run", "--rm",
-            "-v", f"{cls.volume_src}:/data",
-            "alpine:3.20",
-            "sh", "-lc", "mkdir -p /data && echo 'hello' > /data/hello.txt",
-        ])
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.volume_src}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "mkdir -p /data && echo 'hello' > /data/hello.txt",
+            ]
+        )
 
         # databases.csv (unused, but required by CLI)
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
         write_databases_csv(cls.databases_csv, [])
 
-        # Run backup (files should be copied)
         backup_run(
             backups_dir=cls.backups_dir,
             repo_name=cls.repo_name,
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=["dummy-db"],
-            images_no_stop_required=["alpine", "postgres", "mariadb", "mysql"],
+            images_no_stop_required=["alpine:3.20"],
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
@@ -75,20 +79,33 @@ class TestE2EFilesFull(unittest.TestCase):
 
     def test_restore_files_into_new_volume(self) -> None:
         # restore files from volume_src backup into volume_dst
-        run([
-            "baudolo-restore", "files",
-            self.volume_dst, self.hash, self.version,
-            "--backups-dir", self.backups_dir,
-            "--repo-name", self.repo_name,
-            "--source-volume", self.volume_src,
-            "--rsync-image", "ghcr.io/kevinveenbirkenbach/alpine-rsync",
-        ])
+        run(
+            [
+                "baudolo-restore",
+                "files",
+                self.volume_dst,
+                self.hash,
+                self.version,
+                "--backups-dir",
+                self.backups_dir,
+                "--repo-name",
+                self.repo_name,
+                "--source-volume",
+                self.volume_src,
+            ]
+        )
 
-        # verify restored file exists in dst volume
-        p = run([
-            "docker", "run", "--rm",
-            "-v", f"{self.volume_dst}:/data",
-            "alpine:3.20",
-            "sh", "-lc", "cat /data/hello.txt",
-        ])
+        p = run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{self.volume_dst}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "cat /data/hello.txt",
+            ]
+        )
         self.assertEqual((p.stdout or "").strip(), "hello")

tests/e2e/test_e2e_files_no_copy.py

@@ -1,16 +1,16 @@
 import unittest
 
 from .helpers import (
-    backup_run,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
+    run,
     unique,
     write_databases_csv,
-    run,
 )
 
 
@@ -26,47 +26,93 @@ class TestE2EFilesNoCopy(unittest.TestCase):
         cls.repo_name = cls.prefix
 
         cls.volume_src = f"{cls.prefix}-vol-src"
-        cls.volume_dst = f"{cls.prefix}-vol-dst"
-        cls.containers = []
-        cls.volumes = [cls.volume_src, cls.volume_dst]
+        cls.containers: list[str] = []
+        cls.volumes = [cls.volume_src]
 
         run(["docker", "volume", "create", cls.volume_src])
-        run([
-            "docker", "run", "--rm",
-            "-v", f"{cls.volume_src}:/data",
-            "alpine:3.20",
-            "sh", "-lc", "echo 'hello' > /data/hello.txt",
-        ])
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.volume_src}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "echo 'hello' > /data/hello.txt",
+            ]
+        )
 
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
         write_databases_csv(cls.databases_csv, [])
 
-        # dump-only => NO file rsync backups
+        # only-sql => non-DB volumes are STILL backed up as files
         backup_run(
             backups_dir=cls.backups_dir,
             repo_name=cls.repo_name,
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=["dummy-db"],
-            images_no_stop_required=["alpine", "postgres", "mariadb", "mysql"],
-            dump_only=True,
+            images_no_stop_required=["alpine:3.20"],
+            only_sql=True,
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
 
+        # Wipe the volume to ensure restore actually restores something
+        run(["docker", "volume", "rm", "-f", cls.volume_src])
+        run(["docker", "volume", "create", cls.volume_src])
+
     @classmethod
     def tearDownClass(cls) -> None:
         cleanup_docker(containers=cls.containers, volumes=cls.volumes)
 
-    def test_files_backup_not_present(self) -> None:
-        p = backup_path(self.backups_dir, self.repo_name, self.version, self.volume_src) / "files"
-        self.assertFalse(p.exists(), f"Did not expect files backup dir at: {p}")
+    def test_files_backup_present_for_non_db_volume(self) -> None:
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.volume_src)
+            / "files"
+        )
+        self.assertTrue(p.exists(), f"Expected files backup dir at: {p}")
 
-    def test_restore_files_fails_expected(self) -> None:
-        p = run([
-            "baudolo-restore", "files",
-            self.volume_dst, self.hash, self.version,
-            "--backups-dir", self.backups_dir,
-            "--repo-name", self.repo_name,
-        ], check=False)
-        self.assertEqual(p.returncode, 2, f"Expected exitcode 2, got {p.returncode}\nSTDOUT={p.stdout}\nSTDERR={p.stderr}")
+    def test_restore_files_succeeds_and_restores_content(self) -> None:
+        p = run(
+            [
+                "baudolo-restore",
+                "files",
+                self.volume_src,
+                self.hash,
+                self.version,
+                "--backups-dir",
+                self.backups_dir,
+                "--repo-name",
+                self.repo_name,
+            ],
+            check=False,
+        )
+        self.assertEqual(
+            p.returncode,
+            0,
+            f"Expected exitcode 0, got {p.returncode}\nSTDOUT={p.stdout}\nSTDERR={p.stderr}",
+        )
+
+        cp = run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{self.volume_src}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "cat /data/hello.txt",
+            ],
+            capture=True,
+            check=True,
+        )
+        self.assertEqual(
+            cp.stdout.strip(),
+            "hello",
+            f"Unexpected restored content. STDOUT={cp.stdout}\nSTDERR={cp.stderr}",
+        )

tests/e2e/test_e2e_images_no_backup_required_early_skip.py

@@ -0,0 +1,123 @@
+import unittest
+
+from .helpers import (
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    write_databases_csv,
+)
+
+
+class TestE2EImagesNoBackupRequiredEarlySkip(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+
+        cls.prefix = unique("baudolo-e2e-early-skip-no-backup-required")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        # --- Docker resources ---
+        cls.redis_container = f"{cls.prefix}-redis"
+        cls.ignored_volume = f"{cls.prefix}-redis-vol"
+        cls.normal_volume = f"{cls.prefix}-files-vol"
+
+        cls.containers = [cls.redis_container]
+        cls.volumes = [cls.ignored_volume, cls.normal_volume]
+
+        run(["docker", "volume", "create", cls.ignored_volume])
+        run(["docker", "volume", "create", cls.normal_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.redis_container,
+                "-v",
+                f"{cls.ignored_volume}:/data",
+                "redis:alpine",
+            ]
+        )
+
+        # Put deterministic content into the normal volume
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.normal_volume}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "mkdir -p /data && echo 'hello' > /data/hello.txt",
+            ]
+        )
+
+        # databases.csv required by CLI (can be empty)
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(cls.databases_csv, [])
+
+        cmd = [
+            "baudolo",
+            "--compose-dir",
+            cls.compose_dir,
+            "--hard-restart-projects",
+            "mailu",
+            "--repo-name",
+            cls.repo_name,
+            "--databases-csv",
+            cls.databases_csv,
+            "--backups-dir",
+            cls.backups_dir,
+            "--database-containers",
+            "dummy-db",
+            "--images-no-stop-required",
+            "redis:alpine",
+            "--images-no-backup-required",
+            "redis:alpine",
+        ]
+        cp = run(cmd, capture=True, check=True)
+        cls.stdout = cp.stdout or ""
+        cls.stderr = cp.stderr or ""
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_ignored_volume_has_no_backup_directory_at_all(self) -> None:
+        p = backup_path(
+            self.backups_dir,
+            self.repo_name,
+            self.version,
+            self.ignored_volume,
+        )
+        self.assertFalse(
+            p.exists(),
+            f"Expected NO backup directory to be created for ignored volume, but found: {p}",
+        )
+
+    def test_normal_volume_is_still_backed_up(self) -> None:
+        p = (
+            backup_path(
+                self.backups_dir,
+                self.repo_name,
+                self.version,
+                self.normal_volume,
+            )
+            / "files"
+            / "hello.txt"
+        )
+        self.assertTrue(p.is_file(), f"Expected backed up file at: {p}")

tests/e2e/test_e2e_mariadb_anonymous_preemption.py

@@ -0,0 +1,166 @@
+"""
+Bug-repro for: mariadb-dump fails with `ERROR 1045 Access denied for user
+'<u>'@'localhost' (using password: YES)` when only '<u>'@'%' is granted and a
+preempting ''@'localhost' user is present.
+
+The fix forces TCP loopback in baudolo.backup.db so the dump matches the
+'<u>'@'%' grant instead of the socket->localhost auth row.
+
+This file:
+- builds the exact preconditions that triggered the production failure,
+- as a NEGATIVE control, runs a socket-based mariadb-dump (== the old code path)
+  and asserts that it fails with the literal 1045 / @'localhost' error,
+- as a POSITIVE proof, calls backup_database() (where the fix lives) against
+  the same DB container and asserts the dump file is produced and contains the
+  seed data.
+
+Note: the volume-rsync stage of baudolo is intentionally NOT exercised here.
+That stage needs root on /var/lib/docker/volumes, which is provided by the
+DinD wrapper in `make test-e2e` but not by an on-host invocation. The bug we
+are verifying is in the DB-dump stage, so testing backup_database() directly
+keeps the assertion focused and the test runnable both on-host and in DinD.
+"""
+
+import tempfile
+import unittest
+from pathlib import Path
+
+import pandas as pd
+
+from baudolo.backup import db as db_mod
+from baudolo.generation import DUMP_SUFFIX, SQL_DIR
+
+from .helpers import (
+    MARIADB_DATA_DIR,
+    MARIADB_IMAGE,
+    cleanup_docker,
+    require_docker,
+    run,
+    unique,
+    wait_for_mariadb,
+    wait_for_mariadb_sql,
+)
+
+
+class TestE2EMariaDBAnonymousPreemption(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-mariadb-anon")
+        cls.db_container = f"{cls.prefix}-mariadb"
+        cls.db_volume = f"{cls.prefix}-mariadb-vol"
+        cls.containers = [cls.db_container]
+        cls.volumes = [cls.db_volume]
+
+        cls.db_name = "appdb"
+        cls.db_user = "tcponly"
+        cls.db_password = "tcponlypw"
+        cls.root_password = "rootpw"
+
+        run(["docker", "volume", "create", cls.db_volume])
+
+        # Boot WITHOUT MARIADB_USER/MARIADB_PASSWORD/MARIADB_DATABASE so the
+        # entrypoint does not auto-create '<u>'@'%'. We provision the user
+        # explicitly below to mirror the SQL path used by svc-db-mariadb.
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.db_container,
+                "-e",
+                f"MARIADB_ROOT_PASSWORD={cls.root_password}",
+                "-v",
+                f"{cls.db_volume}:{MARIADB_DATA_DIR}",
+                MARIADB_IMAGE,
+            ]
+        )
+
+        wait_for_mariadb(
+            cls.db_container, root_password=cls.root_password, timeout_s=120
+        )
+
+        # Provision: '<u>'@'%' (the app/backup grant) + anonymous ''@'localhost'
+        # (the preemption trigger). Mirrors the production state that produced
+        # `ERROR 1045 ... '<u>'@'localhost' (using password: YES)`.
+        bootstrap_sql = (
+            f"CREATE DATABASE {cls.db_name};"
+            f"CREATE USER '{cls.db_user}'@'%' IDENTIFIED BY '{cls.db_password}';"
+            f"GRANT ALL PRIVILEGES ON {cls.db_name}.* TO '{cls.db_user}'@'%';"
+            f"CREATE USER ''@'localhost' IDENTIFIED BY 'anonpw-not-{cls.db_password}';"
+            "FLUSH PRIVILEGES;"
+            f"CREATE TABLE {cls.db_name}.t (id INT PRIMARY KEY, v VARCHAR(50));"
+            f"INSERT INTO {cls.db_name}.t VALUES (1,'ok');"
+        )
+        run(
+            [
+                "docker",
+                "exec",
+                cls.db_container,
+                "sh",
+                "-lc",
+                f'mariadb -uroot --protocol=socket -e "{bootstrap_sql}"',
+            ]
+        )
+
+        # Sanity: '<u>' can log in over TCP (matches '%'). If THIS fails,
+        # the precondition for the fix to even apply is broken.
+        wait_for_mariadb_sql(
+            cls.db_container, user=cls.db_user, password=cls.db_password, timeout_s=60
+        )
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_negative_control_socket_dump_fails_with_1045(self) -> None:
+        # Reproduces the OLD code path (no -h/--protocol). MUST fail with 1045
+        # under the configured preemption. If this ever starts passing, either
+        # the MariaDB auth semantics changed or the anonymous-user setup did
+        # not take effect — in both cases the positive test below loses its
+        # ability to discriminate "fix works" vs "bug never reproduced".
+        p = run(
+            [
+                "docker",
+                "exec",
+                self.db_container,
+                "sh",
+                "-lc",
+                f"mariadb-dump -u{self.db_user} -p{self.db_password} {self.db_name}",
+            ],
+            capture=True,
+            check=False,
+        )
+        self.assertNotEqual(p.returncode, 0, "socket-based dump unexpectedly succeeded")
+        self.assertIn("1045", (p.stderr or "") + (p.stdout or ""))
+        self.assertIn("@'localhost'", (p.stderr or "") + (p.stdout or ""))
+
+    def test_backup_database_succeeds_with_tcp_fix(self) -> None:
+        # Drives the function where the fix lives. No rsync, no privileged
+        # paths — just the dump that the negative-control proved is failing
+        # under the same preemption setup.
+        with tempfile.TemporaryDirectory() as volume_dir:
+            df = pd.DataFrame(
+                [(self.db_container, self.db_name, self.db_user, self.db_password)],
+                columns=["instance", "database", "username", "password"],
+            )
+            produced = db_mod.backup_database(
+                container=self.db_container,
+                volume_dir=volume_dir,
+                db_type="mariadb",
+                dump_tool="mariadb-dump",
+                databases_df=df,
+                database_containers=[self.db_container],
+            )
+            self.assertTrue(produced, "backup_database did not produce a dump")
+            dump_path = Path(volume_dir) / SQL_DIR / f"{self.db_name}{DUMP_SUFFIX}"
+            self.assertTrue(dump_path.is_file(), f"expected dump at {dump_path}")
+            with dump_path.open(encoding="utf-8", errors="replace") as f:
+                content = f.read()
+            self.assertIn("INSERT INTO", content)
+            self.assertIn("'ok'", content)
+
+
+if __name__ == "__main__":
+    unittest.main(verbosity=2)

tests/e2e/test_e2e_mariadb_full.py

@@ -1,19 +1,20 @@
-# tests/e2e/test_e2e_mariadb_full.py
 import unittest
 
 from .helpers import (
-    backup_run,
+    MARIADB_DATA_DIR,
+    MARIADB_IMAGE,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
-    unique,
-    write_databases_csv,
     run,
+    unique,
     wait_for_mariadb,
     wait_for_mariadb_sql,
+    write_databases_csv,
 )
 
 
@@ -56,16 +57,19 @@ class TestE2EMariaDBFull(unittest.TestCase):
                 "-e",
                 f"MARIADB_PASSWORD={cls.db_password}",
                 "-v",
-                f"{cls.db_volume}:/var/lib/mysql",
-                "mariadb:11",
+                f"{cls.db_volume}:{MARIADB_DATA_DIR}",
+                MARIADB_IMAGE,
             ]
         )
 
         # Liveness + actual SQL login readiness (TCP)
-        wait_for_mariadb(cls.db_container, root_password=cls.root_password, timeout_s=90)
-        wait_for_mariadb_sql(cls.db_container, user=cls.db_user, password=cls.db_password, timeout_s=90)
+        wait_for_mariadb(
+            cls.db_container, root_password=cls.root_password, timeout_s=90
+        )
+        wait_for_mariadb_sql(
+            cls.db_container, user=cls.db_user, password=cls.db_password, timeout_s=90
+        )
 
-        # Create table + data via the dedicated user (TCP)
         run(
             [
                 "docker",
@@ -73,15 +77,20 @@ class TestE2EMariaDBFull(unittest.TestCase):
                 cls.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
-                f"-e \"CREATE TABLE {cls.db_name}.t (id INT PRIMARY KEY, v VARCHAR(50)); "
-                f"INSERT INTO {cls.db_name}.t VALUES (1,'ok');\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
+                    f'-e "CREATE TABLE {cls.db_name}.t (id INT PRIMARY KEY, v VARCHAR(50)); '
+                    f"INSERT INTO {cls.db_name}.t VALUES (1,'ok');\""
+                ),
             ]
         )
 
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
         # IMPORTANT: baudolo backup expects credentials for the DB dump.
-        write_databases_csv(cls.databases_csv, [(cls.db_container, cls.db_name, cls.db_user, cls.db_password)])
+        write_databases_csv(
+            cls.databases_csv,
+            [(cls.db_container, cls.db_name, cls.db_user, cls.db_password)],
+        )
 
         # Backup with file+dump
         backup_run(
@@ -90,7 +99,7 @@ class TestE2EMariaDBFull(unittest.TestCase):
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=[cls.db_container],
-            images_no_stop_required=["mariadb", "mysql", "alpine", "postgres"],
+            images_no_stop_required=[MARIADB_IMAGE],
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
@@ -103,8 +112,10 @@ class TestE2EMariaDBFull(unittest.TestCase):
                 cls.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
-                f"-e \"DROP TABLE {cls.db_name}.t;\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
+                    f'-e "DROP TABLE {cls.db_name}.t;"'
+                ),
             ]
         )
 
@@ -137,7 +148,11 @@ class TestE2EMariaDBFull(unittest.TestCase):
         cleanup_docker(containers=cls.containers, volumes=cls.volumes)
 
     def test_dump_file_exists(self) -> None:
-        p = backup_path(self.backups_dir, self.repo_name, self.version, self.db_volume) / "sql" / f"{self.db_name}.backup.sql"
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.db_volume)
+            / "sql"
+            / f"{self.db_name}.backup.sql"
+        )
         self.assertTrue(p.is_file(), f"Expected dump file at: {p}")
 
     def test_data_restored(self) -> None:
@@ -148,8 +163,10 @@ class TestE2EMariaDBFull(unittest.TestCase):
                 self.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{self.db_user} -p{self.db_password} "
-                f"-N -e \"SELECT v FROM {self.db_name}.t WHERE id=1;\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{self.db_user} -p{self.db_password} "
+                    f'-N -e "SELECT v FROM {self.db_name}.t WHERE id=1;"'
+                ),
             ]
         )
         self.assertEqual((p.stdout or "").strip(), "ok")

tests/e2e/test_e2e_mariadb_no_copy.py

@@ -1,19 +1,20 @@
-# tests/e2e/test_e2e_mariadb_no_copy.py
 import unittest
 
 from .helpers import (
-    backup_run,
+    MARIADB_DATA_DIR,
+    MARIADB_IMAGE,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
-    unique,
-    write_databases_csv,
     run,
+    unique,
     wait_for_mariadb,
     wait_for_mariadb_sql,
+    write_databases_csv,
 )
 
 
@@ -55,15 +56,18 @@ class TestE2EMariaDBNoCopy(unittest.TestCase):
                 "-e",
                 f"MARIADB_PASSWORD={cls.db_password}",
                 "-v",
-                f"{cls.db_volume}:/var/lib/mysql",
-                "mariadb:11",
+                f"{cls.db_volume}:{MARIADB_DATA_DIR}",
+                MARIADB_IMAGE,
             ]
         )
 
-        wait_for_mariadb(cls.db_container, root_password=cls.root_password, timeout_s=90)
-        wait_for_mariadb_sql(cls.db_container, user=cls.db_user, password=cls.db_password, timeout_s=90)
+        wait_for_mariadb(
+            cls.db_container, root_password=cls.root_password, timeout_s=90
+        )
+        wait_for_mariadb_sql(
+            cls.db_container, user=cls.db_user, password=cls.db_password, timeout_s=90
+        )
 
-        # Create table + data (TCP)
         run(
             [
                 "docker",
@@ -71,24 +75,29 @@ class TestE2EMariaDBNoCopy(unittest.TestCase):
                 cls.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
-                f"-e \"CREATE TABLE {cls.db_name}.t (id INT PRIMARY KEY, v VARCHAR(50)); "
-                f"INSERT INTO {cls.db_name}.t VALUES (1,'ok');\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
+                    f'-e "CREATE TABLE {cls.db_name}.t (id INT PRIMARY KEY, v VARCHAR(50)); '
+                    f"INSERT INTO {cls.db_name}.t VALUES (1,'ok');\""
+                ),
             ]
         )
 
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
-        write_databases_csv(cls.databases_csv, [(cls.db_container, cls.db_name, cls.db_user, cls.db_password)])
+        write_databases_csv(
+            cls.databases_csv,
+            [(cls.db_container, cls.db_name, cls.db_user, cls.db_password)],
+        )
 
-        # dump-only => no files
+        # only-sql => no files
         backup_run(
             backups_dir=cls.backups_dir,
             repo_name=cls.repo_name,
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=[cls.db_container],
-            images_no_stop_required=["mariadb", "mysql", "alpine", "postgres"],
-            dump_only=True,
+            images_no_stop_required=[MARIADB_IMAGE],
+            only_sql=True,
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
@@ -101,8 +110,10 @@ class TestE2EMariaDBNoCopy(unittest.TestCase):
                 cls.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
-                f"-e \"DROP TABLE {cls.db_name}.t;\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
+                    f'-e "DROP TABLE {cls.db_name}.t;"'
+                ),
             ]
         )
 
@@ -135,7 +146,10 @@ class TestE2EMariaDBNoCopy(unittest.TestCase):
         cleanup_docker(containers=cls.containers, volumes=cls.volumes)
 
     def test_files_backup_not_present(self) -> None:
-        p = backup_path(self.backups_dir, self.repo_name, self.version, self.db_volume) / "files"
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.db_volume)
+            / "files"
+        )
         self.assertFalse(p.exists(), f"Did not expect files backup dir at: {p}")
 
     def test_data_restored(self) -> None:
@@ -146,8 +160,10 @@ class TestE2EMariaDBNoCopy(unittest.TestCase):
                 self.db_container,
                 "sh",
                 "-lc",
-                f"mariadb -h 127.0.0.1 -u{self.db_user} -p{self.db_password} "
-                f"-N -e \"SELECT v FROM {self.db_name}.t WHERE id=1;\"",
+                (
+                    f"mariadb -h 127.0.0.1 -u{self.db_user} -p{self.db_password} "
+                    f'-N -e "SELECT v FROM {self.db_name}.t WHERE id=1;"'
+                ),
             ]
         )
         self.assertEqual((p.stdout or "").strip(), "ok")

tests/e2e/test_e2e_only_files.py

@@ -0,0 +1,116 @@
+"""--only-files backs a database up as a file tree and asks for no credentials.
+
+The run deliberately passes no --databases-csv at all: a host that only copies
+files has no reason to hold database passwords, and requiring the file would
+make the flag useless there.
+"""
+
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+)
+
+MARKER = "only-files-marker"
+
+
+class TestE2EOnlyFiles(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-only-files")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f"echo '{MARKER}' > {POSTGRES_DATA_DIR}/marker.txt",
+            ]
+        )
+
+        cp = run(
+            [
+                "baudolo",
+                "--compose-dir",
+                cls.compose_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--backups-dir",
+                cls.backups_dir,
+                "--images-no-stop-required",
+                POSTGRES_IMAGE,
+                "--only-files",
+            ],
+            capture=True,
+            check=True,
+        )
+        cls.stdout = cp.stdout or ""
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def _volume_dir(self):
+        return backup_path(
+            self.backups_dir, self.repo_name, self.version, self.pg_volume
+        )
+
+    def test_the_database_volume_is_backed_up_as_files(self) -> None:
+        marker = self._volume_dir() / "files" / "marker.txt"
+        self.assertTrue(marker.is_file(), f"expected a file backup at {marker}")
+        self.assertEqual(marker.read_text(encoding="utf-8").strip(), MARKER)
+
+    def test_no_dump_is_written(self) -> None:
+        sql_dir = self._volume_dir() / "sql"
+        dumps = list(sql_dir.glob("*.sql")) if sql_dir.exists() else []
+        self.assertEqual(dumps, [], f"did not expect any dump, found: {dumps}")
+
+    def test_the_missing_databases_csv_is_not_reported(self) -> None:
+        self.assertNotIn("databases.csv", self.stdout)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_only_sql_fallback_to_files.py

@@ -0,0 +1,204 @@
+import json
+import unittest
+
+from baudolo.generation import FILES_DIR, MANIFEST_FILE, MANIFEST_SCHEMA, SQL_DIR
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+
+class TestE2EOnlySqlFallbackToFiles(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-only-sql-fallback")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.restore_volume = f"{cls.prefix}-restore-vol"
+
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume, cls.restore_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        # Add a deterministic marker file into the volume
+        cls.marker = "only-sql-fallback-marker"
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f"echo '{cls.marker}' > {POSTGRES_DATA_DIR}/marker.txt",
+            ]
+        )
+
+        # databases.csv WITHOUT matching entry for this instance -> should skip dump
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(cls.databases_csv, [])  # empty except header
+
+        # Run baudolo with --only-sql and a DB container present:
+        # Expected: WARNING + FALLBACK to file backup (files/ must exist)
+        cmd = [
+            "baudolo",
+            "--compose-dir",
+            cls.compose_dir,
+            "--hard-restart-projects",
+            "mailu",
+            "--repo-name",
+            cls.repo_name,
+            "--databases-csv",
+            cls.databases_csv,
+            "--backups-dir",
+            cls.backups_dir,
+            "--database-containers",
+            cls.pg_container,
+            "--images-no-stop-required",
+            POSTGRES_IMAGE,
+            "--only-sql",
+        ]
+        cp = run(cmd, capture=True, check=True)
+
+        cls.stdout = cp.stdout or ""
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+        # Restore files into a fresh volume to prove file backup happened
+        run(["docker", "volume", "create", cls.restore_volume])
+        run(
+            [
+                "baudolo-restore",
+                "files",
+                cls.restore_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--source-volume",
+                cls.pg_volume,
+            ]
+        )
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_warns_about_missing_dump_in_dump_only_mode(self) -> None:
+        self.assertIn(
+            "WARNING: only-sql requested but no DB dump was produced",
+            self.stdout,
+            f"Expected warning in baudolo output. STDOUT:\n{self.stdout}",
+        )
+
+    def test_files_backup_exists_due_to_fallback(self) -> None:
+        p = (
+            backup_path(
+                self.backups_dir,
+                self.repo_name,
+                self.version,
+                self.pg_volume,
+            )
+            / "files"
+        )
+        self.assertTrue(p.is_dir(), f"Expected files backup dir at: {p}")
+
+    def test_sql_dump_not_present(self) -> None:
+        # There should be no sql dumps because databases.csv had no matching entry.
+        sql_dir = (
+            backup_path(
+                self.backups_dir,
+                self.repo_name,
+                self.version,
+                self.pg_volume,
+            )
+            / "sql"
+        )
+        # Could exist (dir created) in some edge cases, but should contain no *.sql dumps.
+        if sql_dir.exists():
+            dumps = list(sql_dir.glob("*.sql"))
+            self.assertEqual(
+                len(dumps),
+                0,
+                f"Did not expect SQL dump files, found: {dumps}",
+            )
+
+    def manifest(self) -> dict:
+        generation = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.pg_volume
+        ).parent
+        return json.loads((generation / MANIFEST_FILE).read_text(encoding="utf-8"))
+
+    def test_the_manifest_records_the_volume_as_a_database_left_undumped(self) -> None:
+        """The fallback is invisible in the tree: files/ looks like any copy."""
+        self.assertEqual(
+            self.manifest()["volumes"][self.pg_volume],
+            {"database": True, "dumped": False, "engine": "postgres"},
+        )
+
+    def test_the_manifest_layout_names_where_the_payload_really_landed(self) -> None:
+        layout = self.manifest()["layout"]
+        volume_dir = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.pg_volume
+        )
+        self.assertTrue((volume_dir / layout["files_dir"]).is_dir())
+        self.assertEqual(layout["files_dir"], FILES_DIR)
+        self.assertEqual(layout["sql_dir"], SQL_DIR)
+
+    def test_the_manifest_states_a_schema_a_reader_can_check(self) -> None:
+        self.assertEqual(self.manifest()["schema"], MANIFEST_SCHEMA)
+
+    def test_restored_files_contain_marker(self) -> None:
+        p = run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{self.restore_volume}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "cat /data/marker.txt",
+            ]
+        )
+        self.assertEqual((p.stdout or "").strip(), self.marker)

tests/e2e/test_e2e_only_sql_mixed_run.py

@@ -0,0 +1,202 @@
+import json
+import unittest
+
+from baudolo.generation import MANIFEST_FILE
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+
+class TestE2EOnlySqlMixedRun(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-only-sql-mixed-run")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        # --- Volumes ---
+        cls.db_volume = f"{cls.prefix}-vol-db"
+        cls.files_volume = f"{cls.prefix}-vol-files"
+
+        # Track for cleanup
+        cls.containers: list[str] = []
+        cls.volumes = [cls.db_volume, cls.files_volume]
+
+        run(["docker", "volume", "create", cls.db_volume])
+        run(["docker", "volume", "create", cls.files_volume])
+
+        # Put a marker into the non-db volume
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.files_volume}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "echo 'hello-non-db' > /data/hello.txt",
+            ]
+        )
+
+        # --- Start Postgres container using the DB volume ---
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.containers.append(cls.pg_container)
+
+        cls.pg_password = "postgres"
+        cls.pg_db = "testdb"
+        cls.pg_user = "postgres"
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                f"POSTGRES_PASSWORD={cls.pg_password}",
+                "-v",
+                f"{cls.db_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        # Create deterministic content in DB so dump is non-empty
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -c "CREATE DATABASE {cls.pg_db};" || true',
+            ],
+            check=True,
+        )
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                (
+                    f"psql -U postgres -d {cls.pg_db} -c "
+                    '"CREATE TABLE IF NOT EXISTS t (id INT PRIMARY KEY, v TEXT);'
+                    "INSERT INTO t(id,v) VALUES (1,'hello-db') "
+                    'ON CONFLICT (id) DO UPDATE SET v=EXCLUDED.v;"'
+                ),
+            ],
+            check=True,
+        )
+
+        # databases.csv with an entry => dump should succeed
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv,
+            [(cls.pg_container, cls.pg_db, cls.pg_user, cls.pg_password)],
+        )
+
+        cmd = [
+            "baudolo",
+            "--compose-dir",
+            cls.compose_dir,
+            "--databases-csv",
+            cls.databases_csv,
+            "--database-containers",
+            cls.pg_container,
+            "--images-no-stop-required",
+            POSTGRES_IMAGE,
+            "--only-sql",
+            "--backups-dir",
+            cls.backups_dir,
+            "--repo-name",
+            cls.repo_name,
+        ]
+        cp = run(cmd, capture=True, check=True)
+        cls.stdout = cp.stdout
+        cls.stderr = cp.stderr
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_db_volume_has_dump_and_no_files_dir(self) -> None:
+        base = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.db_volume
+        )
+
+        dumps = base / "sql"
+        files = base / "files"
+
+        self.assertTrue(dumps.exists(), f"Expected dumps dir for DB volume at: {dumps}")
+        self.assertFalse(
+            files.exists(),
+            f"Did not expect files dir for DB volume when dump succeeded at: {files}",
+        )
+
+        # Optional: at least one dump file exists
+        dump_files = list(dumps.glob("*.sql")) + list(dumps.glob("*.sql.gz"))
+        self.assertTrue(
+            dump_files,
+            f"Expected at least one SQL dump file in {dumps}, found none.",
+        )
+
+    def test_non_db_volume_has_files_dir(self) -> None:
+        base = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.files_volume
+        )
+        files = base / "files"
+        self.assertTrue(
+            files.exists(),
+            f"Expected files dir for non-DB volume at: {files}",
+        )
+
+    def test_only_sql_does_not_disable_non_db_files_backup(self) -> None:
+        # Regression guard: even with --only-sql, non-DB volumes must still be backed up as files
+        base = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.files_volume
+        )
+        self.assertTrue(
+            (base / "files").exists(),
+            f"Expected non-DB volume files backup to exist at: {base / 'files'}",
+        )
+
+    def manifest(self) -> dict:
+        generation = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.db_volume
+        ).parent
+        return json.loads((generation / MANIFEST_FILE).read_text(encoding="utf-8"))
+
+    def test_the_manifest_records_the_dumped_volume_as_dumped(self) -> None:
+        self.assertEqual(
+            self.manifest()["volumes"][self.db_volume],
+            {"database": True, "dumped": True, "engine": "postgres"},
+        )
+
+    def test_the_manifest_records_the_plain_volume_as_no_database(self) -> None:
+        self.assertEqual(
+            self.manifest()["volumes"][self.files_volume],
+            {"database": False, "dumped": False, "engine": None},
+        )

tests/e2e/test_e2e_postgres_cluster_restore.py

@@ -0,0 +1,215 @@
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+# One statement per entry: psql wraps a multi-statement -c in a transaction,
+# and CREATE DATABASE is forbidden inside one.
+SEED_SQL = (
+    "CREATE ROLE app LOGIN PASSWORD 'apppw'",
+    "CREATE DATABASE first OWNER app",
+    "CREATE DATABASE second OWNER app",
+)
+SIBLING_SQL = (
+    "CREATE ROLE neighbour LOGIN PASSWORD 'neighbourpw'",
+    "CREATE DATABASE sibling OWNER neighbour",
+)
+SIBLING_PAYLOAD = "CREATE TABLE t (v text); INSERT INTO t VALUES ('sibling-payload');"
+FIRST_SQL = "CREATE TABLE t (v text); INSERT INTO t VALUES ('first-payload');"
+SECOND_SQL = "CREATE TABLE t (v text); INSERT INTO t VALUES ('second-payload');"
+
+
+class TestE2EPostgresClusterRestore(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-pg-cluster")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres")
+
+        for statement in SEED_SQL:
+            cls._psql("postgres", statement)
+        cls._psql("first", FIRST_SQL)
+        cls._psql("second", SECOND_SQL)
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "*", "postgres", "pgpw")]
+        )
+
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.pg_container],
+            images_no_stop_required=[POSTGRES_IMAGE],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+        cls.dump = (
+            backup_path(cls.backups_dir, cls.repo_name, cls.version, cls.pg_volume)
+            / "sql"
+            / f"{cls.pg_container}.cluster.backup.sql"
+        )
+
+        run(
+            [
+                "baudolo-restore",
+                "cluster",
+                cls.pg_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.pg_container,
+                "--instance",
+                cls.pg_container,
+                "--db-user",
+                "postgres",
+                "--db-password",
+                "pgpw",
+                "--empty",
+            ]
+        )
+
+        for statement in SIBLING_SQL:
+            cls._psql("postgres", statement)
+        cls._psql("sibling", SIBLING_PAYLOAD)
+        cls.refused = run(cls._restore_argv(), check=False)
+
+    @classmethod
+    def _restore_argv(cls) -> list:
+        return [
+            "baudolo-restore",
+            "cluster",
+            cls.pg_volume,
+            cls.hash,
+            cls.version,
+            "--backups-dir",
+            cls.backups_dir,
+            "--repo-name",
+            cls.repo_name,
+            "--container",
+            cls.pg_container,
+            "--instance",
+            cls.pg_container,
+            "--db-user",
+            "postgres",
+            "--db-password",
+            "pgpw",
+            "--empty",
+        ]
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    @classmethod
+    def _psql(cls, database: str, sql: str) -> str:
+        p = run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -d {database} -t -A -c "{sql}"',
+            ]
+        )
+        return (p.stdout or "").strip()
+
+    def test_the_backup_wrote_a_cluster_dump(self) -> None:
+        self.assertTrue(self.dump.is_file(), f"no cluster dump at {self.dump}")
+
+    def test_the_preclean_really_dropped_a_populated_cluster(self) -> None:
+        self.assertEqual(self._psql("first", "SELECT v FROM t"), "first-payload")
+
+    def test_a_second_empty_is_refused_once_a_foreign_database_exists(self) -> None:
+        self.assertNotEqual(self.refused.returncode, 0, self.refused.stdout)
+        self.assertIn("sibling", self.refused.stderr)
+
+    def test_the_refusal_left_the_foreign_database_alone(self) -> None:
+        self.assertEqual(self._psql("sibling", "SELECT v FROM t"), "sibling-payload")
+
+    def test_the_refusal_dropped_nothing_of_its_own(self) -> None:
+        self.assertEqual(self._psql("first", "SELECT v FROM t"), "first-payload")
+
+    def test_both_databases_are_back(self) -> None:
+        listed = self._psql(
+            "postgres",
+            "SELECT datname FROM pg_database WHERE datname IN ('first','second') ORDER BY 1",
+        )
+        self.assertEqual(listed.split(), ["first", "second"])
+
+    def test_each_database_carries_its_own_payload(self) -> None:
+        self.assertEqual(self._psql("first", "SELECT v FROM t"), "first-payload")
+        self.assertEqual(self._psql("second", "SELECT v FROM t"), "second-payload")
+
+    def test_the_superusers_own_create_was_filtered(self) -> None:
+        self.assertEqual(
+            self._psql(
+                "postgres", "SELECT rolsuper FROM pg_roles WHERE rolname = 'postgres'"
+            ),
+            "t",
+        )
+
+    def test_the_owning_role_is_back(self) -> None:
+        self.assertEqual(
+            self._psql(
+                "postgres", "SELECT rolname FROM pg_roles WHERE rolname = 'app'"
+            ),
+            "app",
+        )
+
+    def test_ownership_survived(self) -> None:
+        self.assertEqual(
+            self._psql(
+                "postgres",
+                "SELECT pg_get_userbyid(datdba) FROM pg_database WHERE datname = 'first'",
+            ),
+            "app",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_postgres_empty_drop_hard.py

@@ -0,0 +1,185 @@
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+# The scenario the --empty pre-clean must survive: a non-public user schema
+# plus one object of every class the discovery SELECT enumerates. Restore
+# --empty runs against the still-populated DB (no wipe), so the pre-clean must
+# drop discourse_functions too or the dump's CREATE SCHEMA aborts the replay
+# under ON_ERROR_STOP; the old public-only DROP left it and broke discourse.
+# The f()/f(int) pair reproduces discourse's overload abort ("function name
+# is not unique") and english_stem_nostop reproduces taiga's text search
+# dictionary abort (duplicate pg_ts_dict_dictname_index).
+SCENARIO_SQL = (
+    "CREATE SCHEMA discourse_functions;"
+    "CREATE TABLE discourse_functions.helper (id int);"
+    "INSERT INTO discourse_functions.helper VALUES (1);"
+    "CREATE TABLE public.t (id int primary key, v text);"
+    "INSERT INTO public.t VALUES (1, 'ok');"
+    "CREATE VIEW public.t_view AS SELECT * FROM public.t;"
+    "CREATE SEQUENCE public.s;"
+    "CREATE TYPE public.mood AS ENUM ('ok', 'bad');"
+    "CREATE FUNCTION public.f() RETURNS int LANGUAGE sql AS 'SELECT 1';"
+    "CREATE FUNCTION public.f(i int) RETURNS int LANGUAGE sql AS 'SELECT i';"
+    "CREATE COLLATION public.c (locale = 'C');"
+    "CREATE TEXT SEARCH DICTIONARY public.english_stem_nostop"
+    " (Template = snowball, Language = english);"
+    "CREATE TEXT SEARCH CONFIGURATION public.english_nostop (COPY = english);"
+    "ALTER TEXT SEARCH CONFIGURATION public.english_nostop"
+    " ALTER MAPPING FOR asciiword WITH public.english_stem_nostop;"
+)
+
+
+class TestE2EPostgresEmptyDropHard(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-postgres-empty-drop-hard")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -d appdb -v ON_ERROR_STOP=1 -c "{SCENARIO_SQL}"',
+            ]
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")]
+        )
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.pg_container],
+            images_no_stop_required=[POSTGRES_IMAGE],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+        # No wipe: restore --empty must pre-clean the fully-populated DB
+        # (incl. the non-public schema) before replaying the dump.
+        run(
+            [
+                "baudolo-restore",
+                "postgres",
+                cls.pg_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.pg_container,
+                "--db-name",
+                "appdb",
+                "--db-user",
+                "postgres",
+                "--db-password",
+                "pgpw",
+                "--empty",
+            ]
+        )
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def _scalar(self, sql: str) -> str:
+        p = run(
+            [
+                "docker",
+                "exec",
+                self.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -d appdb -t -A -c "{sql}"',
+            ]
+        )
+        return (p.stdout or "").strip()
+
+    def test_public_data_restored(self) -> None:
+        self.assertEqual(self._scalar("SELECT v FROM public.t WHERE id=1;"), "ok")
+
+    def test_view_restored(self) -> None:
+        self.assertEqual(self._scalar("SELECT count(*) FROM public.t_view;"), "1")
+
+    def test_non_public_schema_restored(self) -> None:
+        self.assertEqual(
+            self._scalar(
+                "SELECT count(*) FROM pg_namespace WHERE nspname='discourse_functions';"
+            ),
+            "1",
+        )
+
+    def test_overloaded_functions_restored_once_each(self) -> None:
+        self.assertEqual(
+            self._scalar("SELECT count(*) FROM pg_proc WHERE proname='f';"), "2"
+        )
+        self.assertEqual(self._scalar("SELECT public.f(41) + public.f();"), "42")
+
+    def test_text_search_dictionary_restored_once(self) -> None:
+        self.assertEqual(
+            self._scalar(
+                "SELECT count(*) FROM pg_ts_dict WHERE dictname='english_stem_nostop';"
+            ),
+            "1",
+        )
+        self.assertEqual(
+            self._scalar(
+                "SELECT count(*) FROM pg_ts_config WHERE cfgname='english_nostop';"
+            ),
+            "1",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_postgres_full.py

@@ -1,18 +1,19 @@
-# tests/e2e/test_e2e_postgres_full.py
 import unittest
 
 from .helpers import (
-    backup_run,
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
-    unique,
-    write_databases_csv,
     run,
+    unique,
     wait_for_postgres,
+    write_databases_csv,
 )
 
 
@@ -33,26 +34,41 @@ class TestE2EPostgresFull(unittest.TestCase):
 
         run(["docker", "volume", "create", cls.pg_volume])
 
-        run([
-            "docker", "run", "-d",
-            "--name", cls.pg_container,
-            "-e", "POSTGRES_PASSWORD=pgpw",
-            "-e", "POSTGRES_DB=appdb",
-            "-e", "POSTGRES_USER=postgres",
-            "-v", f"{cls.pg_volume}:/var/lib/postgresql/data",
-            "postgres:16",
-        ])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
         wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
 
-        # Create a table + data
-        run([
-            "docker", "exec", cls.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -c \"CREATE TABLE t (id int primary key, v text); INSERT INTO t VALUES (1,'ok');\"",
-        ])
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                "psql -U postgres -d appdb -c \"CREATE TABLE t (id int primary key, v text); INSERT INTO t VALUES (1,'ok');\"",
+            ]
+        )
 
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
-        write_databases_csv(cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")])
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")]
+        )
 
         backup_run(
             backups_dir=cls.backups_dir,
@@ -60,43 +76,68 @@ class TestE2EPostgresFull(unittest.TestCase):
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=[cls.pg_container],
-            images_no_stop_required=["postgres", "mariadb", "mysql", "alpine"],
+            images_no_stop_required=[POSTGRES_IMAGE],
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
 
         # Wipe schema
-        run([
-            "docker", "exec", cls.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -c \"DROP TABLE t;\"",
-        ])
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                'psql -U postgres -d appdb -c "DROP TABLE t;"',
+            ]
+        )
 
         # Restore
-        run([
-            "baudolo-restore", "postgres",
-            cls.pg_volume, cls.hash, cls.version,
-            "--backups-dir", cls.backups_dir,
-            "--repo-name", cls.repo_name,
-            "--container", cls.pg_container,
-            "--db-name", "appdb",
-            "--db-user", "postgres",
-            "--db-password", "pgpw",
-            "--empty",
-        ])
+        run(
+            [
+                "baudolo-restore",
+                "postgres",
+                cls.pg_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.pg_container,
+                "--db-name",
+                "appdb",
+                "--db-user",
+                "postgres",
+                "--db-password",
+                "pgpw",
+                "--empty",
+            ]
+        )
 
     @classmethod
     def tearDownClass(cls) -> None:
         cleanup_docker(containers=cls.containers, volumes=cls.volumes)
 
     def test_dump_file_exists(self) -> None:
-        p = backup_path(self.backups_dir, self.repo_name, self.version, self.pg_volume) / "sql" / "appdb.backup.sql"
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.pg_volume)
+            / "sql"
+            / "appdb.backup.sql"
+        )
         self.assertTrue(p.is_file(), f"Expected dump file at: {p}")
 
     def test_data_restored(self) -> None:
-        p = run([
-            "docker", "exec", self.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -t -c \"SELECT v FROM t WHERE id=1;\"",
-        ])
+        p = run(
+            [
+                "docker",
+                "exec",
+                self.pg_container,
+                "sh",
+                "-lc",
+                'psql -U postgres -d appdb -t -c "SELECT v FROM t WHERE id=1;"',
+            ]
+        )
         self.assertEqual((p.stdout or "").strip(), "ok")

tests/e2e/test_e2e_postgres_no_copy.py

@@ -1,18 +1,19 @@
-# tests/e2e/test_e2e_postgres_no_copy.py
 import unittest
 
 from .helpers import (
-    backup_run,
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
     backup_path,
+    backup_run,
     cleanup_docker,
     create_minimal_compose_dir,
     ensure_empty_dir,
     latest_version_dir,
     require_docker,
-    unique,
-    write_databases_csv,
     run,
+    unique,
     wait_for_postgres,
+    write_databases_csv,
 )
 
 
@@ -32,25 +33,41 @@ class TestE2EPostgresNoCopy(unittest.TestCase):
         cls.volumes = [cls.pg_volume]
 
         run(["docker", "volume", "create", cls.pg_volume])
-        run([
-            "docker", "run", "-d",
-            "--name", cls.pg_container,
-            "-e", "POSTGRES_PASSWORD=pgpw",
-            "-e", "POSTGRES_DB=appdb",
-            "-e", "POSTGRES_USER=postgres",
-            "-v", f"{cls.pg_volume}:/var/lib/postgresql/data",
-            "postgres:16",
-        ])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
         wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
 
-        run([
-            "docker", "exec", cls.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -c \"CREATE TABLE t (id int primary key, v text); INSERT INTO t VALUES (1,'ok');\"",
-        ])
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                "psql -U postgres -d appdb -c \"CREATE TABLE t (id int primary key, v text); INSERT INTO t VALUES (1,'ok');\"",
+            ]
+        )
 
         cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
-        write_databases_csv(cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")])
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")]
+        )
 
         backup_run(
             backups_dir=cls.backups_dir,
@@ -58,42 +75,66 @@ class TestE2EPostgresNoCopy(unittest.TestCase):
             compose_dir=cls.compose_dir,
             databases_csv=cls.databases_csv,
             database_containers=[cls.pg_container],
-            images_no_stop_required=["postgres", "mariadb", "mysql", "alpine"],
-            dump_only=True,
+            images_no_stop_required=[POSTGRES_IMAGE],
+            only_sql=True,
         )
 
         cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
 
-        run([
-            "docker", "exec", cls.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -c \"DROP TABLE t;\"",
-        ])
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                'psql -U postgres -d appdb -c "DROP TABLE t;"',
+            ]
+        )
 
-        run([
-            "baudolo-restore", "postgres",
-            cls.pg_volume, cls.hash, cls.version,
-            "--backups-dir", cls.backups_dir,
-            "--repo-name", cls.repo_name,
-            "--container", cls.pg_container,
-            "--db-name", "appdb",
-            "--db-user", "postgres",
-            "--db-password", "pgpw",
-            "--empty",
-        ])
+        run(
+            [
+                "baudolo-restore",
+                "postgres",
+                cls.pg_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.pg_container,
+                "--db-name",
+                "appdb",
+                "--db-user",
+                "postgres",
+                "--db-password",
+                "pgpw",
+                "--empty",
+            ]
+        )
 
     @classmethod
     def tearDownClass(cls) -> None:
         cleanup_docker(containers=cls.containers, volumes=cls.volumes)
 
     def test_files_backup_not_present(self) -> None:
-        p = backup_path(self.backups_dir, self.repo_name, self.version, self.pg_volume) / "files"
+        p = (
+            backup_path(self.backups_dir, self.repo_name, self.version, self.pg_volume)
+            / "files"
+        )
         self.assertFalse(p.exists(), f"Did not expect files backup dir at: {p}")
 
     def test_data_restored(self) -> None:
-        p = run([
-            "docker", "exec", self.pg_container,
-            "sh", "-lc",
-            "psql -U postgres -d appdb -t -c \"SELECT v FROM t WHERE id=1;\"",
-        ])
+        p = run(
+            [
+                "docker",
+                "exec",
+                self.pg_container,
+                "sh",
+                "-lc",
+                'psql -U postgres -d appdb -t -c "SELECT v FROM t WHERE id=1;"',
+            ]
+        )
         self.assertEqual((p.stdout or "").strip(), "ok")

tests/e2e/test_e2e_postgres_password_required.py

@@ -0,0 +1,159 @@
+"""An engine whose loopback auth really demands a password.
+
+Every other Postgres scenario runs stock postgres:alpine, whose generated
+pg_hba grants trust on 127.0.0.1 and ::1 - so `pg_dump -h localhost` never
+needs the password and a dump succeeds whether or not baudolo hands one to the
+container. This module makes the password mandatory, which is what a dedicated
+engine on a real host does.
+"""
+
+import unittest
+from pathlib import Path
+
+from baudolo.generation import CLUSTER_SUFFIX, DUMP_SUFFIX, FILES_DIR, SQL_DIR
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+
+class TestE2EPostgresPasswordRequired(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-pg-password-required")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.containers = [cls.pg_container]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                # The entrypoint evals this into its initdb call, so the host
+                # lines of pg_hba demand scram while the local socket stays
+                # trust - the entrypoint's own init and the seeding below keep
+                # working, and only a TCP connection needs the password.
+                "-e",
+                "POSTGRES_INITDB_ARGS=--auth-host=scram-sha-256",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                (
+                    'psql -U postgres -d appdb -c "CREATE TABLE t (id int primary '
+                    "key, v text); INSERT INTO t VALUES (1,'ok');\""
+                ),
+            ],
+            check=True,
+        )
+
+        cls.unauthenticated = run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                "pg_dump -U postgres -d appdb -h localhost",
+            ],
+            capture=True,
+            check=False,
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv,
+            [
+                (cls.pg_container, "appdb", "postgres", "pgpw"),
+                (cls.pg_container, "*", "postgres", "pgpw"),
+            ],
+        )
+
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.pg_container],
+            images_no_stop_required=[POSTGRES_IMAGE],
+            only_sql=True,
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def volume_dir(self) -> Path:
+        return backup_path(
+            self.backups_dir, self.repo_name, self.version, self.pg_volume
+        )
+
+    def test_a_dump_without_the_password_is_refused_by_the_server(self) -> None:
+        """Without this the module is vacuous: a pg_hba still saying trust would
+        let a baudolo that forwards nothing pass just as well."""
+        self.assertNotEqual(self.unauthenticated.returncode, 0)
+        self.assertIn("no password supplied", self.unauthenticated.stderr or "")
+
+    def test_the_configured_database_was_dumped(self) -> None:
+        dump = self.volume_dir() / SQL_DIR / f"appdb{DUMP_SUFFIX}"
+        self.assertTrue(dump.is_file(), f"expected a dump at {dump}")
+        self.assertIn("Dumped by pg_dump", dump.read_text(encoding="utf-8"))
+
+    def test_the_dump_carries_the_payload(self) -> None:
+        """pg_dump emits table data as COPY ... FROM stdin, so the row reads as
+        tab-separated values rather than as an INSERT literal."""
+        dump = self.volume_dir() / SQL_DIR / f"appdb{DUMP_SUFFIX}"
+        self.assertIn("COPY public.t (id, v) FROM stdin;", dump.read_text("utf-8"))
+        self.assertIn("1\tok", dump.read_text(encoding="utf-8"))
+
+    def test_the_cluster_row_was_dumped_too(self) -> None:
+        cluster = self.volume_dir() / SQL_DIR / f"{self.pg_container}{CLUSTER_SUFFIX}"
+        self.assertTrue(cluster.is_file(), f"expected a cluster dump at {cluster}")
+        self.assertIn("CREATE DATABASE", cluster.read_text(encoding="utf-8"))
+
+    def test_only_sql_left_no_file_copy_behind(self) -> None:
+        self.assertFalse((self.volume_dir() / FILES_DIR).exists())
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_postgres_single_transaction_live_writer.py

@@ -0,0 +1,184 @@
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+# The discourse restore-drill race: `restore --empty` replays the dump into a
+# LIVE database while a background writer keeps touching a primary-key row
+# (discourse's mini_scheduler upserts scheduler_stats(id=1)). Without a
+# single-transaction replay, the pre-clean drops the table, the replay recreates
+# it and auto-commits, the writer wins the gap and inserts id=1, and the dump's
+# COPY of the same id then aborts with a duplicate-key violation under
+# ON_ERROR_STOP -> the whole restore fails. The --single-transaction replay keeps
+# the recreated table invisible until commit, so the writer can never insert the
+# racing row and the restore completes. A wide filler table makes the COPY slow
+# enough that the non-transactional variant loses the race deterministically.
+SEED_SQL = (
+    "CREATE TABLE public.scheduler_stats (id int primary key, v text);"
+    "INSERT INTO public.scheduler_stats VALUES (1, 'from-dump');"
+    "CREATE TABLE public.filler (id serial primary key, blob text);"
+    "INSERT INTO public.filler (blob)"
+    " SELECT repeat('x', 512) FROM generate_series(1, 100000);"
+)
+
+WRITER_LOOP = (
+    "while true; do "
+    "psql -h 127.0.0.1 -U postgres -d appdb "
+    "-c \"INSERT INTO public.scheduler_stats(id, v) VALUES (1, 'live') "
+    'ON CONFLICT (id) DO NOTHING;" >/dev/null 2>&1; '
+    "done"
+)
+
+
+class TestE2EPostgresSingleTransactionLiveWriter(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-pg-single-txn")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.pg_volume = f"{cls.prefix}-pg-vol"
+        cls.writer = f"{cls.prefix}-writer"
+        cls.containers = [cls.pg_container, cls.writer]
+        cls.volumes = [cls.pg_volume]
+
+        run(["docker", "volume", "create", cls.pg_volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                "POSTGRES_PASSWORD=pgpw",
+                "-e",
+                "POSTGRES_DB=appdb",
+                "-e",
+                "POSTGRES_USER=postgres",
+                "-v",
+                f"{cls.pg_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -d appdb -v ON_ERROR_STOP=1 -c "{SEED_SQL}"',
+            ]
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv, [(cls.pg_container, "appdb", "postgres", "pgpw")]
+        )
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.pg_container],
+            images_no_stop_required=[POSTGRES_IMAGE],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.writer,
+                "--network",
+                f"container:{cls.pg_container}",
+                "-e",
+                "PGPASSWORD=pgpw",
+                POSTGRES_IMAGE,
+                "sh",
+                "-lc",
+                WRITER_LOOP,
+            ]
+        )
+
+        cls.restore = run(
+            [
+                "baudolo-restore",
+                "postgres",
+                cls.pg_volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.pg_container,
+                "--db-name",
+                "appdb",
+                "--db-user",
+                "postgres",
+                "--db-password",
+                "pgpw",
+                "--empty",
+            ],
+            capture=True,
+            check=False,
+        )
+
+        run(["docker", "rm", "-f", cls.writer], capture=True, check=False)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def _scalar(self, sql: str) -> str:
+        p = run(
+            [
+                "docker",
+                "exec",
+                self.pg_container,
+                "sh",
+                "-lc",
+                f'psql -U postgres -d appdb -t -A -c "{sql}"',
+            ]
+        )
+        return (p.stdout or "").strip()
+
+    def test_restore_survived_the_live_writer(self) -> None:
+        self.assertEqual(
+            self.restore.returncode,
+            0,
+            f"restore aborted (duplicate-key race not contained):\n{self.restore.stderr}",
+        )
+
+    def test_primary_key_row_restored(self) -> None:
+        self.assertEqual(
+            self._scalar("SELECT count(*) FROM public.scheduler_stats WHERE id=1;"),
+            "1",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_restore_files_backing_store.py

@@ -0,0 +1,120 @@
+"""Restoring files into a volume that has a backing store of its own.
+
+Docker keeps the same ``/var/lib/docker/volumes/<name>/_data`` path for such a
+volume and mounts the real storage over it only while a container holds it.
+Writing there unmounted lands in the empty directory underneath, is hidden by
+the next mount, and rsync reports success - so the restore has to refuse.
+"""
+
+import unittest
+from pathlib import Path
+
+from .helpers import (
+    backup_path,
+    cleanup_docker,
+    ensure_empty_dir,
+    machine_hash,
+    require_docker,
+    run,
+    unique,
+)
+
+MARKER = "restored-payload"
+VERSION = "20260817000000"
+
+
+def mountpoint_of(volume: str) -> Path:
+    return Path("/var/lib/docker/volumes") / volume / "_data"
+
+
+def contents(directory: Path) -> list[str]:
+    return sorted(p.name for p in directory.iterdir()) if directory.is_dir() else []
+
+
+class TestE2ERestoreFilesBackingStore(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-backing")
+        cls.repo_name = cls.prefix
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        cls.backing = Path(f"/tmp/{cls.prefix}/backing")
+        ensure_empty_dir(cls.backups_dir)
+        ensure_empty_dir(str(cls.backing))
+
+        cls.bound_volume = f"{cls.prefix}-bound"
+        cls.plain_volume = f"{cls.prefix}-plain"
+        cls.volumes = [cls.bound_volume, cls.plain_volume]
+
+        for volume in cls.volumes:
+            files = (
+                backup_path(cls.backups_dir, cls.repo_name, VERSION, volume) / "files"
+            )
+            files.mkdir(parents=True, exist_ok=True)
+            (files / "marker.txt").write_text(MARKER, encoding="utf-8")
+
+        run(
+            [
+                "docker",
+                "volume",
+                "create",
+                "--driver",
+                "local",
+                "--opt",
+                "type=none",
+                "--opt",
+                "o=bind",
+                "--opt",
+                f"device={cls.backing}",
+                cls.bound_volume,
+            ]
+        )
+        run(["docker", "volume", "create", cls.plain_volume])
+
+        cls.refused = cls.restore(cls.bound_volume)
+        cls.accepted = cls.restore(cls.plain_volume)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=[], volumes=cls.volumes)
+
+    @classmethod
+    def restore(cls, volume: str):
+        return run(
+            [
+                "baudolo-restore",
+                "files",
+                volume,
+                machine_hash(),
+                VERSION,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+            ],
+            check=False,
+        )
+
+    def test_a_volume_with_its_own_backing_store_is_refused(self) -> None:
+        self.assertEqual(self.refused.returncode, 2, self.refused.stdout)
+        self.assertIn("backing store of its own", self.refused.stderr)
+
+    def test_nothing_was_written_into_the_backing_store(self) -> None:
+        self.assertEqual(contents(self.backing), [])
+
+    def test_nothing_was_written_under_the_mount_either(self) -> None:
+        self.assertEqual(
+            contents(mountpoint_of(self.bound_volume)),
+            [],
+            "the copy landed in the directory the next mount hides",
+        )
+
+    def test_a_plain_volume_is_still_restored(self) -> None:
+        self.assertEqual(self.accepted.returncode, 0, self.accepted.stderr)
+        restored = mountpoint_of(self.plain_volume) / "marker.txt"
+        self.assertTrue(restored.is_file(), f"{restored} missing")
+        self.assertEqual(restored.read_text(encoding="utf-8"), MARKER)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_seed_star_and_db_entries_backup_postgres.py

@@ -0,0 +1,230 @@
+import unittest
+
+from .helpers import (
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_postgres,
+)
+
+
+class TestE2ESeedStarAndDbEntriesBackupPostgres(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+
+        cls.prefix = unique("baudolo-e2e-seed-star-and-db")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        # --- Volumes ---
+        cls.db_volume = f"{cls.prefix}-vol-db"
+        cls.files_volume = f"{cls.prefix}-vol-files"
+        cls.volumes = [cls.db_volume, cls.files_volume]
+
+        run(["docker", "volume", "create", cls.db_volume])
+        run(["docker", "volume", "create", cls.files_volume])
+
+        # Put a marker into the non-db volume
+        cls.marker = "hello-non-db-seed-star"
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.files_volume}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                f"echo '{cls.marker}' > /data/hello.txt",
+            ]
+        )
+
+        # --- Start Postgres container using the DB volume ---
+        cls.pg_container = f"{cls.prefix}-pg"
+        cls.containers = [cls.pg_container]
+
+        cls.pg_password = "postgres"
+        cls.pg_user = "postgres"
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.pg_container,
+                "-e",
+                f"POSTGRES_PASSWORD={cls.pg_password}",
+                "-v",
+                f"{cls.db_volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.pg_container, user="postgres", timeout_s=90)
+
+        # Create two DBs and deterministic content, so pg_dumpall is meaningful
+        cls.pg_db1 = "testdb1"
+        cls.pg_db2 = "testdb2"
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                (
+                    f'psql -U {cls.pg_user} -c "CREATE DATABASE {cls.pg_db1};" || true; '
+                    f'psql -U {cls.pg_user} -c "CREATE DATABASE {cls.pg_db2};" || true; '
+                ),
+            ],
+            check=True,
+        )
+
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                (
+                    f"psql -U {cls.pg_user} -d {cls.pg_db1} -c "
+                    '"CREATE TABLE IF NOT EXISTS t (id INT PRIMARY KEY, v TEXT);'
+                    "INSERT INTO t(id,v) VALUES (1,'hello-db1') "
+                    'ON CONFLICT (id) DO UPDATE SET v=EXCLUDED.v;"'
+                ),
+            ],
+            check=True,
+        )
+        run(
+            [
+                "docker",
+                "exec",
+                cls.pg_container,
+                "sh",
+                "-lc",
+                (
+                    f"psql -U {cls.pg_user} -d {cls.pg_db2} -c "
+                    '"CREATE TABLE IF NOT EXISTS t (id INT PRIMARY KEY, v TEXT);'
+                    "INSERT INTO t(id,v) VALUES (1,'hello-db2') "
+                    'ON CONFLICT (id) DO UPDATE SET v=EXCLUDED.v;"'
+                ),
+            ],
+            check=True,
+        )
+
+        # --- Seed databases.csv using CLI (star + concrete db) ---
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+
+        # IMPORTANT: because we pass --database-containers <container>,
+        # get_instance() will use the container name as instance key.
+        instance = cls.pg_container
+
+        # Seed star entry (pg_dumpall)
+        run(
+            [
+                "baudolo-seed",
+                cls.databases_csv,
+                instance,
+                "*",
+                cls.pg_user,
+                cls.pg_password,
+            ]
+        )
+
+        # Seed concrete DB entry (pg_dump)
+        run(
+            [
+                "baudolo-seed",
+                cls.databases_csv,
+                instance,
+                cls.pg_db1,
+                cls.pg_user,
+                cls.pg_password,
+            ]
+        )
+
+        cmd = [
+            "baudolo",
+            "--compose-dir",
+            cls.compose_dir,
+            "--databases-csv",
+            cls.databases_csv,
+            "--database-containers",
+            cls.pg_container,
+            "--images-no-stop-required",
+            POSTGRES_IMAGE,
+            "--only-sql",
+            "--backups-dir",
+            cls.backups_dir,
+            "--repo-name",
+            cls.repo_name,
+        ]
+        cp = run(cmd, capture=True, check=True)
+        cls.stdout = cp.stdout or ""
+        cls.stderr = cp.stderr or ""
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_db_volume_has_cluster_dump_and_concrete_db_dump_and_no_files(self) -> None:
+        base = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.db_volume
+        )
+        sql_dir = base / "sql"
+        files_dir = base / "files"
+
+        self.assertTrue(sql_dir.exists(), f"Expected sql dir at: {sql_dir}")
+        self.assertFalse(
+            files_dir.exists(),
+            f"Did not expect files dir for DB volume when only-sql succeeded: {files_dir}",
+        )
+
+        # Cluster dump file produced by '*' entry
+        cluster = sql_dir / f"{self.pg_container}.cluster.backup.sql"
+        self.assertTrue(cluster.is_file(), f"Expected cluster dump file at: {cluster}")
+
+        # Concrete DB dump produced by normal entry
+        db1 = sql_dir / f"{self.pg_db1}.backup.sql"
+        self.assertTrue(db1.is_file(), f"Expected db dump file at: {db1}")
+
+        # Basic sanity: cluster dump usually contains CREATE DATABASE statements
+        txt = cluster.read_text(encoding="utf-8", errors="ignore")
+        self.assertIn(
+            "CREATE DATABASE",
+            txt,
+            "Expected cluster dump to contain CREATE DATABASE statements",
+        )
+
+    def test_non_db_volume_still_has_files_backup(self) -> None:
+        base = backup_path(
+            self.backups_dir, self.repo_name, self.version, self.files_volume
+        )
+        files_dir = base / "files"
+
+        self.assertTrue(
+            files_dir.exists(), f"Expected files dir for non-DB volume at: {files_dir}"
+        )
+
+        marker = files_dir / "hello.txt"
+        self.assertTrue(marker.is_file(), f"Expected marker file at: {marker}")
+        self.assertEqual(
+            marker.read_text(encoding="utf-8").strip(),
+            self.marker,
+        )

tests/e2e/test_e2e_snapshot.py

@@ -0,0 +1,169 @@
+"""Snapshot capture against real filesystems.
+
+Loop devices are only available to a privileged container, so each case builds
+its filesystem inside one and drives snapshot_driver.py there. A filesystem
+without snapshot support must fail loudly rather than degrade to a live copy,
+which is the property that makes the mode safe to offer at all.
+"""
+
+from __future__ import annotations
+
+import os
+import shutil
+import unittest
+from pathlib import Path
+
+from .helpers import require_docker, run, unique
+
+REPO_SRC = Path(__file__).resolve().parents[2] / "src"
+DRIVER = Path(__file__).resolve().parent / "snapshot_driver.py"
+FAITHFUL_DRIVER = Path(__file__).resolve().parent / "faithful_driver.py"
+IMAGE = "alpine:3.20"
+PACKAGES = "apk add -q btrfs-progs e2fsprogs zfs python3 util-linux"
+ATTACH = (
+    "LOOP=$(losetup -f | awk '{print $1}') "
+    '&& { [ -b "$LOOP" ] || mknod "$LOOP" b 7 "${LOOP#/dev/loop}"; }; '
+    'losetup "$LOOP" /img'
+)
+LOOP_FS = {
+    "btrfs": "btrfs subvolume create /subject/docker >/dev/null",
+    "ext4": "mkdir -p /subject/docker",
+}
+# A container carries no /lib/modules, so modprobe fails even on a loaded module.
+ZFS_READY = "{ [ -c /dev/zfs ] || modprobe zfs 2>/dev/null; }; [ -c /dev/zfs ]"
+
+
+def mount_script(fstype: str) -> str:
+    """Build a filesystem on a loop device and carve out the snapshot subject."""
+    if fstype == "zfs":
+        return (
+            f"{PACKAGES} && {ZFS_READY} && truncate -s 400M /img "
+            "&& zpool create -m none baudolo /img "
+            "&& zfs create -o mountpoint=/subject/docker baudolo/docker"
+        )
+    return (
+        f"{PACKAGES} && truncate -s 400M /img && mkfs.{fstype} -q /img "
+        f'&& mkdir -p /subject && {ATTACH} && mount -t {fstype} "$LOOP" /subject '
+        f"&& {LOOP_FS[fstype]}"
+    )
+
+
+def zfs_usable() -> bool:
+    """Whether this host's kernel can serve zfs to a privileged container."""
+    proc = run(
+        [
+            "docker",
+            "run",
+            "--rm",
+            "--privileged",
+            IMAGE,
+            "sh",
+            "-lc",
+            f"apk add -q zfs >/dev/null 2>&1 && {ZFS_READY}",
+        ],
+        capture=True,
+        check=False,
+    )
+    return proc.returncode == 0
+
+
+def required(fstype: str) -> bool:
+    """Whether this run must cover ``fstype`` instead of skipping it.
+
+    CI sets E2E_REQUIRE_FILESYSTEMS so a missing kernel module fails the build
+    rather than passing it with a filesystem silently untested.
+    """
+    demanded = os.environ.get("E2E_REQUIRE_FILESYSTEMS", "")
+    return fstype in demanded.replace(",", " ").split()
+
+
+def stage(driver: Path) -> Path:
+    """Copy source and driver under /tmp, the only path the DinD daemon shares."""
+    staged = Path("/tmp") / unique("baudolo-e2e-snapshot")
+    shutil.copytree(REPO_SRC, staged / "src")
+    shutil.copy(driver, staged / "driver.py")
+    return staged
+
+
+def drive(fstype: str, arguments: str, *, driver: Path = DRIVER) -> str:
+    staged = stage(driver)
+    script = f"set -e; {mount_script(fstype)}; python3 /driver.py {arguments}"
+    try:
+        proc = run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "--privileged",
+                "--name",
+                staged.name,
+                "-v",
+                f"{staged / 'src'}:/src:ro",
+                "-v",
+                f"{staged / 'driver.py'}:/driver.py:ro",
+                IMAGE,
+                "sh",
+                "-lc",
+                script,
+            ],
+            capture=True,
+            check=False,
+        )
+    finally:
+        shutil.rmtree(staged, ignore_errors=True)
+    if proc.returncode != 0:
+        raise AssertionError(
+            f"{fstype} driver failed on {arguments}:\n{proc.stdout}\n{proc.stderr}"
+        )
+    return proc.stdout
+
+
+class TestE2ESnapshot(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+
+    def assert_freezes(self, fstype: str) -> None:
+        output = drive(fstype, f"{fstype} /subject/docker supported")
+        self.assertIn("PASS the snapshot exposes the volume", output)
+        self.assertIn("PASS a later write does not reach the snapshot", output)
+        self.assertIn("PASS the snapshot is removed afterwards", output)
+        self.assertIn("ALL OK", output)
+
+    def test_btrfs_snapshot_freezes_the_volume(self) -> None:
+        self.assert_freezes("btrfs")
+
+    def test_zfs_snapshot_freezes_the_volume(self) -> None:
+        if not zfs_usable():
+            if required("zfs"):
+                self.fail(
+                    "E2E_REQUIRE_FILESYSTEMS demands zfs, but this kernel provides no "
+                    "zfs module; load it before running the suite"
+                )
+            self.skipTest(
+                "this kernel provides no zfs module, so no pool can be created"
+            )
+        self.assert_freezes("zfs")
+
+    def test_ext4_has_no_snapshot_and_says_so(self) -> None:
+        output = drive("ext4", "btrfs /subject/docker unsupported")
+        self.assertIn("PASS refused loudly", output)
+
+    def test_an_unknown_kind_is_refused_before_touching_the_filesystem(self) -> None:
+        output = drive("ext4", "lvm /subject/docker unsupported")
+        self.assertIn("PASS refused loudly", output)
+
+    def test_a_volume_with_its_own_storage_is_copied_live_not_from_the_snapshot(
+        self,
+    ) -> None:
+        output = drive("btrfs", "/subject/docker", driver=FAITHFUL_DRIVER)
+        self.assertIn(
+            "PASS the snapshot shows the other volume as an empty directory", output
+        )
+        self.assertIn("PASS the other volume degrades to live", output)
+        self.assertIn("PASS the plain volume is read from the snapshot", output)
+        self.assertIn("ALL OK", output)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_snapshot_db.py

@@ -0,0 +1,112 @@
+"""A live database survives being captured from a snapshot.
+
+The database is written to while the snapshot is taken and keeps writing
+afterwards, so the copy can only be a point in time - never a clean shutdown.
+A second server is then started on that copy: it must recover on its own and
+still hold every row committed before the snapshot.
+"""
+
+from __future__ import annotations
+
+import shutil
+import unittest
+from pathlib import Path
+
+from .helpers import require_docker, run, unique
+
+REPO_SRC = Path(__file__).resolve().parents[2] / "src"
+DRIVER = Path(__file__).resolve().parent / "snapshot_db_driver.py"
+IMAGE = "alpine:3.20"
+SOCKET = "/tmp/live.sock"
+RESTORED_SOCKET = "/tmp/restored.sock"
+DATADIR = "/subject/docker/volumes/mariadb_data/_data"
+RESTORED = "/restored"
+
+SCRIPT = f"""set -e
+apk add -q btrfs-progs util-linux python3 mariadb mariadb-client rsync
+truncate -s 900M /img
+mkfs.btrfs -q /img
+mkdir -p /subject
+LOOP=$(losetup -f | awk '{{print $1}}')
+{{ [ -b "$LOOP" ] || mknod "$LOOP" b 7 "${{LOOP#/dev/loop}}"; }}
+losetup "$LOOP" /img
+mount -t btrfs "$LOOP" /subject
+btrfs subvolume create /subject/docker >/dev/null
+mkdir -p {DATADIR}
+
+mariadb-install-db --user=root --datadir={DATADIR} >/dev/null 2>&1
+mariadbd --user=root --datadir={DATADIR} --socket={SOCKET} --skip-networking &
+for i in $(seq 1 60); do mariadb-admin --socket={SOCKET} ping >/dev/null 2>&1 && break; sleep 1; done
+
+mariadb --socket={SOCKET} -e "CREATE DATABASE demo;
+  CREATE TABLE demo.t (id INT PRIMARY KEY, v VARCHAR(32)) ENGINE=InnoDB;
+  INSERT INTO demo.t VALUES (1,'committed'),(2,'committed');"
+
+mariadb --socket={SOCKET} -e "INSERT INTO demo.t VALUES (3,'committed');"
+python3 /driver.py
+mariadb --socket={SOCKET} -e "INSERT INTO demo.t VALUES (4,'after-snapshot');"
+
+mkdir -p {RESTORED}
+rsync -a /backups/20260731/mariadb_data/files/ {RESTORED}/
+mariadbd --user=root --datadir={RESTORED} --socket={RESTORED_SOCKET} --skip-networking &
+for i in $(seq 1 60); do mariadb-admin --socket={RESTORED_SOCKET} ping >/dev/null 2>&1 && break; sleep 1; done
+
+echo "RESTORED_ROWS=$(mariadb --socket={RESTORED_SOCKET} -N -B -e 'SELECT COUNT(*) FROM demo.t;')"
+echo "RESTORED_AFTER=$(mariadb --socket={RESTORED_SOCKET} -N -B -e \\
+  "SELECT COUNT(*) FROM demo.t WHERE v='after-snapshot';")"
+echo DB_OK
+"""
+
+
+class TestE2ESnapshotDatabase(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        staged = Path("/tmp") / unique("baudolo-e2e-snapshot-db")
+        shutil.copytree(REPO_SRC, staged / "src")
+        shutil.copy(DRIVER, staged / "driver.py")
+        try:
+            proc = run(
+                [
+                    "docker",
+                    "run",
+                    "--rm",
+                    "--privileged",
+                    "--name",
+                    staged.name,
+                    "-v",
+                    f"{staged / 'src'}:/src:ro",
+                    "-v",
+                    f"{staged / 'driver.py'}:/driver.py:ro",
+                    IMAGE,
+                    "sh",
+                    "-lc",
+                    SCRIPT,
+                ],
+                capture=True,
+                check=False,
+            )
+        finally:
+            shutil.rmtree(staged, ignore_errors=True)
+        cls.output = proc.stdout + proc.stderr
+        cls.returncode = proc.returncode
+
+    def test_the_run_completed(self) -> None:
+        self.assertEqual(self.returncode, 0, self.output)
+        self.assertIn("DB_OK", self.output)
+
+    def test_the_backup_came_from_the_snapshot(self) -> None:
+        self.assertIn("SNAPSHOT COPY DONE", self.output)
+
+    def test_the_restored_server_recovered_on_its_own(self) -> None:
+        self.assertIn("RESTORED_ROWS=3", self.output)
+
+    def test_writes_after_the_snapshot_are_absent(self) -> None:
+        self.assertIn("RESTORED_AFTER=0", self.output)
+
+    def test_the_copy_was_an_unclean_state_the_engine_had_to_repair(self) -> None:
+        self.assertRegex(self.output, r"(?i)crash recovery|rolling back|log sequence")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_swarm_task_skip.py

@@ -0,0 +1,194 @@
+# Reproduces the swarm flake fixed on this branch: baudolo used to stop a
+# swarm task container around the volume file backup because its image was
+# not whitelisted; the orchestrator immediately replaced the stopped task and
+# the later `docker start` failed on the detached overlay network, killing
+# the backup run. With the fix the task container is skipped (backed up hot):
+# the backup succeeds, the very same container instance keeps running, and
+# the service never has to replace a task.
+import time
+import unittest
+
+from .helpers import (
+    backup_path,
+    backup_run,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    write_databases_csv,
+)
+
+
+def _swarm_state() -> str:
+    return run(
+        ["docker", "info", "--format", "{{.Swarm.LocalNodeState}}"]
+    ).stdout.strip()
+
+
+def _task_container_id(service: str, timeout_s: int = 60) -> str:
+    deadline = time.time() + timeout_s
+    while time.time() < deadline:
+        out = run(
+            [
+                "docker",
+                "ps",
+                "--filter",
+                f"label=com.docker.swarm.service.name={service}",
+                "--format",
+                "{{.ID}}",
+            ]
+        ).stdout.strip()
+        if out:
+            return out.splitlines()[0]
+        time.sleep(2)
+    raise RuntimeError(f"No running task container for service {service}")
+
+
+def _started_at(container_id: str) -> str:
+    return run(
+        ["docker", "inspect", "--format", "{{.State.StartedAt}}", container_id]
+    ).stdout.strip()
+
+
+class TestE2ESwarmTaskSkip(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-swarm-skip")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.swarm_initted = False
+        if _swarm_state() != "active":
+            run(["docker", "swarm", "init", "--advertise-addr", "127.0.0.1"])
+            cls.swarm_initted = True
+
+        cls.volume = f"{cls.prefix}-vol"
+        cls.service = f"{cls.prefix}-svc"
+        cls.volumes = [cls.volume]
+
+        run(["docker", "volume", "create", cls.volume])
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.volume}:/data",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "echo 'swarm-payload' > /data/payload.txt",
+            ]
+        )
+
+        run(
+            [
+                "docker",
+                "service",
+                "create",
+                "--name",
+                cls.service,
+                "--replicas",
+                "1",
+                "--mount",
+                f"type=volume,source={cls.volume},target=/data",
+                "alpine:3.20",
+                "sleep",
+                "3600",
+            ]
+        )
+        cls.task_cid = _task_container_id(cls.service)
+        cls.task_started_at = _started_at(cls.task_cid)
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(cls.databases_csv, [])
+
+        # Whitelist that matches nothing: on main this forces a stop of every
+        # container at the volume, i.e. exactly the flake; on this branch the
+        # swarm task must be skipped instead. (An empty list would leave the
+        # --images-no-stop-required flag without arguments and argparse-fail.)
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=["dummy-db"],
+            images_no_stop_required=["image-that-matches-nothing"],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        run(["docker", "service", "rm", cls.service], check=False)
+        deadline = time.time() + 30
+        while time.time() < deadline:
+            out = run(
+                [
+                    "docker",
+                    "ps",
+                    "-aq",
+                    "--filter",
+                    f"label=com.docker.swarm.service.name={cls.service}",
+                ],
+                check=False,
+            ).stdout.strip()
+            if not out:
+                break
+            time.sleep(2)
+        for v in cls.volumes:
+            run(["docker", "volume", "rm", "-f", v], check=False)
+        if cls.swarm_initted:
+            run(["docker", "swarm", "leave", "--force"], check=False)
+
+    def test_volume_backed_up_hot(self) -> None:
+        p = (
+            backup_path(
+                self.backups_dir,
+                self.repo_name,
+                self.version,
+                self.volume,
+            )
+            / "files"
+            / "payload.txt"
+        )
+        self.assertTrue(p.is_file(), f"Expected backed up file at: {p}")
+
+    def test_task_container_never_stopped(self) -> None:
+        out = run(
+            ["docker", "ps", "-q", "--no-trunc", "--filter", f"id={self.task_cid}"]
+        ).stdout.strip()
+        self.assertTrue(
+            out.startswith(self.task_cid) or self.task_cid.startswith(out.strip()[:12]),
+            f"Task container {self.task_cid} is no longer running",
+        )
+        self.assertEqual(
+            self.task_started_at,
+            _started_at(self.task_cid),
+            "Task container was restarted during the backup",
+        )
+
+    def test_service_never_replaced_the_task(self) -> None:
+        states = run(
+            [
+                "docker",
+                "service",
+                "ps",
+                self.service,
+                "--format",
+                "{{.DesiredState}} {{.CurrentState}}",
+            ]
+        ).stdout.strip()
+        lines = [line for line in states.splitlines() if line.strip()]
+        self.assertEqual(
+            len(lines), 1, f"Service task history shows replacements:\n{states}"
+        )
+        self.assertIn("Running", lines[0])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_version_gate.py

@@ -0,0 +1,283 @@
+"""A dump from a newer engine must be refused before --empty destroys anything.
+
+The pre-clean and the replay are two separate sessions with no rollback across
+them, so a dump the engine cannot parse leaves an emptied database behind. The
+decisive assertion here is not the non-zero exit - it is that the payload is
+still readable afterwards.
+"""
+
+import re
+import unittest
+from pathlib import Path
+
+from .helpers import (
+    MARIADB_DATA_DIR,
+    MARIADB_IMAGE,
+    POSTGRES_DATA_DIR,
+    POSTGRES_IMAGE,
+    backup_path,
+    backup_run,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    wait_for_mariadb,
+    wait_for_mariadb_sql,
+    wait_for_postgres,
+    write_databases_csv,
+)
+
+PAYLOAD = "gate-payload"
+FUTURE = "99.0"
+
+
+def rewrite_version(dump: Path, pattern: str, version: str) -> str:
+    """Make the dump claim ``version``; return what it claimed before."""
+    text = dump.read_text(encoding="utf-8", errors="replace")
+    found = re.search(pattern, text)
+    if not found:
+        raise AssertionError(f"{dump} carries no version header matching {pattern}")
+    claimed = found.group(1)
+    dump.write_text(
+        text.replace(found.group(0), found.group(0).replace(claimed, version), 1),
+        encoding="utf-8",
+    )
+    return claimed
+
+
+class GateCase:
+    """Drive one engine through refusal, escape hatch and truthful replay."""
+
+    engine = ""
+    pattern = ""
+
+    @classmethod
+    def restore(cls, *extra: str):
+        return run(
+            [
+                "baudolo-restore",
+                cls.engine,
+                cls.volume,
+                cls.hash,
+                cls.version,
+                "--backups-dir",
+                cls.backups_dir,
+                "--repo-name",
+                cls.repo_name,
+                "--container",
+                cls.container,
+                "--db-name",
+                cls.db_name,
+                "--db-user",
+                cls.db_user,
+                "--db-password",
+                cls.db_password,
+                "--empty",
+                *extra,
+            ],
+            check=False,
+        )
+
+    @classmethod
+    def prepare(cls) -> None:
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(
+            cls.databases_csv,
+            [(cls.container, cls.db_name, cls.db_user, cls.db_password)],
+        )
+        backup_run(
+            backups_dir=cls.backups_dir,
+            repo_name=cls.repo_name,
+            compose_dir=cls.compose_dir,
+            databases_csv=cls.databases_csv,
+            database_containers=[cls.container],
+            images_no_stop_required=[cls.image],
+        )
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+        cls.dump = (
+            backup_path(cls.backups_dir, cls.repo_name, cls.version, cls.volume)
+            / "sql"
+            / f"{cls.db_name}.backup.sql"
+        )
+
+        cls.truthful_version = rewrite_version(cls.dump, cls.pattern, FUTURE)
+        cls.refused = cls.restore()
+        cls.payload_after_refusal = cls.read_payload()
+
+        cls.forced = cls.restore("--no-version-check")
+        cls.payload_after_force = cls.read_payload()
+
+        rewrite_version(cls.dump, cls.pattern, cls.truthful_version)
+        cls.replayed = cls.restore()
+        cls.payload_after_replay = cls.read_payload()
+
+    def test_the_dump_states_the_engine_it_came_from(self) -> None:
+        self.assertRegex(self.truthful_version, r"^\d+")
+
+    def test_a_newer_dump_is_refused(self) -> None:
+        self.assertNotEqual(self.refused.returncode, 0, self.refused.stdout)
+
+    def test_the_refusal_names_the_version_it_refused(self) -> None:
+        self.assertIn(FUTURE, self.refused.stderr)
+        self.assertIn("older engine", self.refused.stderr)
+
+    def test_the_refusal_left_the_data_untouched(self) -> None:
+        self.assertEqual(
+            self.payload_after_refusal,
+            PAYLOAD,
+            "--empty pre-cleaned before the version was checked",
+        )
+
+    def test_the_escape_hatch_replays_anyway(self) -> None:
+        self.assertEqual(self.forced.returncode, 0, self.forced.stderr)
+        self.assertEqual(self.payload_after_force, PAYLOAD)
+
+    def test_a_truthful_dump_replays(self) -> None:
+        self.assertEqual(self.replayed.returncode, 0, self.replayed.stderr)
+        self.assertEqual(self.payload_after_replay, PAYLOAD)
+
+
+class TestE2EPostgresVersionGate(GateCase, unittest.TestCase):
+    engine = "postgres"
+    pattern = r"-- Dumped from database version (\S+)"
+
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-pg-gate")
+        cls.container = f"{cls.prefix}-pg"
+        cls.volume = f"{cls.prefix}-pg-vol"
+        cls.image = POSTGRES_IMAGE
+        cls.db_name = "appdb"
+        cls.db_user = "postgres"
+        cls.db_password = "pgpw"
+
+        run(["docker", "volume", "create", cls.volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.container,
+                "-e",
+                f"POSTGRES_PASSWORD={cls.db_password}",
+                "-v",
+                f"{cls.volume}:{POSTGRES_DATA_DIR}",
+                POSTGRES_IMAGE,
+            ]
+        )
+        wait_for_postgres(cls.container, user=cls.db_user)
+        cls.sql("postgres", f"CREATE DATABASE {cls.db_name}")
+        cls.sql(
+            cls.db_name,
+            f"CREATE TABLE t (v text); INSERT INTO t VALUES ('{PAYLOAD}');",
+        )
+        cls.prepare()
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=[cls.container], volumes=[cls.volume])
+
+    @classmethod
+    def sql(cls, database: str, statement: str) -> str:
+        p = run(
+            [
+                "docker",
+                "exec",
+                cls.container,
+                "sh",
+                "-lc",
+                f'psql -U {cls.db_user} -d {database} -t -A -c "{statement}"',
+            ],
+            check=False,
+        )
+        return (p.stdout or "").strip()
+
+    @classmethod
+    def read_payload(cls) -> str:
+        return cls.sql(cls.db_name, "SELECT v FROM t")
+
+
+class TestE2EMariadbVersionGate(GateCase, unittest.TestCase):
+    engine = "mariadb"
+    pattern = r"-- Server version\s+(\S+)"
+
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+        cls.prefix = unique("baudolo-e2e-mdb-gate")
+        cls.container = f"{cls.prefix}-mdb"
+        cls.volume = f"{cls.prefix}-mdb-vol"
+        cls.image = MARIADB_IMAGE
+        cls.db_name = "appdb"
+        cls.db_user = "test"
+        cls.db_password = "testpw"
+
+        run(["docker", "volume", "create", cls.volume])
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.container,
+                "-e",
+                "MARIADB_ROOT_PASSWORD=rootpw",
+                "-e",
+                f"MARIADB_DATABASE={cls.db_name}",
+                "-e",
+                f"MARIADB_USER={cls.db_user}",
+                "-e",
+                f"MARIADB_PASSWORD={cls.db_password}",
+                "-v",
+                f"{cls.volume}:{MARIADB_DATA_DIR}",
+                MARIADB_IMAGE,
+            ]
+        )
+        wait_for_mariadb(cls.container, root_password="rootpw", timeout_s=90)
+        wait_for_mariadb_sql(
+            cls.container, user=cls.db_user, password=cls.db_password, timeout_s=90
+        )
+        cls.sql(
+            f"CREATE TABLE {cls.db_name}.t (v VARCHAR(50)); "
+            f"INSERT INTO {cls.db_name}.t VALUES ('{PAYLOAD}');"
+        )
+        cls.prepare()
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=[cls.container], volumes=[cls.volume])
+
+    @classmethod
+    def sql(cls, statement: str) -> str:
+        p = run(
+            [
+                "docker",
+                "exec",
+                cls.container,
+                "sh",
+                "-lc",
+                (
+                    f"mariadb -h 127.0.0.1 -u{cls.db_user} -p{cls.db_password} "
+                    f'-N -B -e "{statement}"'
+                ),
+            ],
+            check=False,
+        )
+        return (p.stdout or "").strip()
+
+    @classmethod
+    def read_payload(cls) -> str:
+        return cls.sql(f"SELECT v FROM {cls.db_name}.t")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/e2e/test_e2e_volumes_no_backup_required_early_skip.py

@@ -0,0 +1,125 @@
+import unittest
+
+from .helpers import (
+    backup_path,
+    cleanup_docker,
+    create_minimal_compose_dir,
+    ensure_empty_dir,
+    latest_version_dir,
+    require_docker,
+    run,
+    unique,
+    write_databases_csv,
+)
+
+
+class TestE2EVolumesNoBackupRequiredEarlySkip(unittest.TestCase):
+    """Both volumes hang off the same container, so an image-level exclusion
+    could only drop both. Only the named one may disappear."""
+
+    @classmethod
+    def setUpClass(cls) -> None:
+        require_docker()
+
+        cls.prefix = unique("baudolo-e2e-early-skip-no-backup-volume")
+        cls.backups_dir = f"/tmp/{cls.prefix}/Backups"
+        ensure_empty_dir(cls.backups_dir)
+
+        cls.compose_dir = create_minimal_compose_dir(f"/tmp/{cls.prefix}")
+        cls.repo_name = cls.prefix
+
+        cls.container = f"{cls.prefix}-app"
+        cls.excluded_volume = f"{cls.prefix}-derived-vol"
+        cls.kept_volume = f"{cls.prefix}-state-vol"
+
+        cls.containers = [cls.container]
+        cls.volumes = [cls.excluded_volume, cls.kept_volume]
+
+        run(["docker", "volume", "create", cls.excluded_volume])
+        run(["docker", "volume", "create", cls.kept_volume])
+
+        run(
+            [
+                "docker",
+                "run",
+                "--rm",
+                "-v",
+                f"{cls.excluded_volume}:/derived",
+                "-v",
+                f"{cls.kept_volume}:/state",
+                "alpine:3.20",
+                "sh",
+                "-lc",
+                "echo derived > /derived/derived.txt && echo state > /state/state.txt",
+            ]
+        )
+
+        run(
+            [
+                "docker",
+                "run",
+                "-d",
+                "--name",
+                cls.container,
+                "-v",
+                f"{cls.excluded_volume}:/derived",
+                "-v",
+                f"{cls.kept_volume}:/state",
+                "alpine:3.20",
+                "sleep",
+                "600",
+            ]
+        )
+
+        cls.databases_csv = f"/tmp/{cls.prefix}/databases.csv"
+        write_databases_csv(cls.databases_csv, [])
+
+        cmd = [
+            "baudolo",
+            "--compose-dir",
+            cls.compose_dir,
+            "--repo-name",
+            cls.repo_name,
+            "--databases-csv",
+            cls.databases_csv,
+            "--backups-dir",
+            cls.backups_dir,
+            "--images-no-stop-required",
+            "alpine:3.20",
+            "--volumes-no-backup-required",
+            cls.excluded_volume,
+        ]
+        cp = run(cmd, capture=True, check=True)
+        cls.stdout = cp.stdout or ""
+        cls.stderr = cp.stderr or ""
+
+        cls.hash, cls.version = latest_version_dir(cls.backups_dir, cls.repo_name)
+
+    @classmethod
+    def tearDownClass(cls) -> None:
+        cleanup_docker(containers=cls.containers, volumes=cls.volumes)
+
+    def test_excluded_volume_has_no_backup_directory_at_all(self) -> None:
+        p = backup_path(
+            self.backups_dir,
+            self.repo_name,
+            self.version,
+            self.excluded_volume,
+        )
+        self.assertFalse(
+            p.exists(),
+            f"Expected NO backup directory for the excluded volume, but found: {p}",
+        )
+
+    def test_sibling_volume_of_the_same_container_is_still_backed_up(self) -> None:
+        p = (
+            backup_path(
+                self.backups_dir,
+                self.repo_name,
+                self.version,
+                self.kept_volume,
+            )
+            / "files"
+            / "state.txt"
+        )
+        self.assertTrue(p.is_file(), f"Expected backed up file at: {p}")

tests/integration/test_seed_integration.py

@@ -6,8 +6,48 @@ import unittest
 from pathlib import Path
 
 
-def run_seed(csv_path: Path, instance: str, database: str, username: str, password: str = "") -> subprocess.CompletedProcess:
-    # Run the real CLI module (integration-style).
+def run_seed(
+    csv_path: Path, instance: str, database: str, username: str, password: str
+) -> subprocess.CompletedProcess:
+    """
+    Run the real CLI module (E2E-style) using subprocess.
+
+    Seed contract (current):
+    - database must be "*" or a valid name (non-empty, matches allowed charset)
+    - password is required
+    - entry is keyed by (instance, database); username/password get updated
+    """
+    cp = subprocess.run(
+        [
+            sys.executable,
+            "-m",
+            "baudolo.seed",
+            str(csv_path),
+            instance,
+            database,
+            username,
+            password,
+        ],
+        text=True,
+        capture_output=True,
+        check=False,
+    )
+    if cp.returncode != 0:
+        raise AssertionError(
+            "seed command failed unexpectedly.\n"
+            f"returncode: {cp.returncode}\n"
+            f"stdout:\n{cp.stdout}\n"
+            f"stderr:\n{cp.stderr}\n"
+        )
+    return cp
+
+
+def run_seed_expect_fail(
+    csv_path: Path, instance: str, database: str, username: str, password: str
+) -> subprocess.CompletedProcess:
+    """
+    Same as run_seed, but expects non-zero exit. Returns CompletedProcess for inspection.
+    """
     return subprocess.run(
         [
             sys.executable,
@@ -21,7 +61,7 @@ def run_seed(csv_path: Path, instance: str, database: str, username: str, passwo
         ],
         text=True,
         capture_output=True,
-        check=True,
+        check=False,
     )
 
 
@@ -31,6 +71,10 @@ def read_csv_semicolon(path: Path) -> list[dict]:
         return list(reader)
 
 
+def read_text(path: Path) -> str:
+    return path.read_text(encoding="utf-8")
+
+
 class TestSeedIntegration(unittest.TestCase):
     def test_creates_file_and_adds_entry_when_missing(self) -> None:
         with tempfile.TemporaryDirectory() as td:
@@ -39,7 +83,7 @@ class TestSeedIntegration(unittest.TestCase):
 
             cp = run_seed(p, "docker.test", "appdb", "alice", "secret")
 
-            self.assertEqual(cp.returncode, 0, cp.stderr)
+            self.assertEqual(cp.returncode, 0)
             self.assertTrue(p.exists())
 
             rows = read_csv_semicolon(p)
@@ -49,40 +93,121 @@ class TestSeedIntegration(unittest.TestCase):
             self.assertEqual(rows[0]["username"], "alice")
             self.assertEqual(rows[0]["password"], "secret")
 
-    def test_replaces_existing_entry_same_keys(self) -> None:
+    def test_replaces_existing_entry_same_instance_and_database_updates_username_and_password(
+        self,
+    ) -> None:
+        """
+        Replacement semantics:
+        - Key is (instance, database)
+        - username/password are updated in-place
+        """
         with tempfile.TemporaryDirectory() as td:
             p = Path(td) / "databases.csv"
 
-            # First add
             run_seed(p, "docker.test", "appdb", "alice", "oldpw")
             rows = read_csv_semicolon(p)
             self.assertEqual(len(rows), 1)
+            self.assertEqual(rows[0]["username"], "alice")
             self.assertEqual(rows[0]["password"], "oldpw")
 
-            # Replace (same instance+database+username)
-            run_seed(p, "docker.test", "appdb", "alice", "newpw")
+            run_seed(p, "docker.test", "appdb", "bob", "newpw")
             rows = read_csv_semicolon(p)
 
             self.assertEqual(len(rows), 1, "Expected replacement, not a duplicate row")
             self.assertEqual(rows[0]["instance"], "docker.test")
             self.assertEqual(rows[0]["database"], "appdb")
-            self.assertEqual(rows[0]["username"], "alice")
+            self.assertEqual(rows[0]["username"], "bob")
             self.assertEqual(rows[0]["password"], "newpw")
 
-    def test_database_empty_string_matches_existing_empty_database(self) -> None:
+    def test_allows_star_database_for_dump_all(self) -> None:
         with tempfile.TemporaryDirectory() as td:
             p = Path(td) / "databases.csv"
 
-            # Add with empty database
-            run_seed(p, "docker.test", "", "alice", "pw1")
+            cp = run_seed(p, "bigbluebutton", "*", "postgres", "pw")
+            self.assertEqual(cp.returncode, 0)
+
             rows = read_csv_semicolon(p)
             self.assertEqual(len(rows), 1)
-            self.assertEqual(rows[0]["database"], "")
+            self.assertEqual(rows[0]["instance"], "bigbluebutton")
+            self.assertEqual(rows[0]["database"], "*")
+            self.assertEqual(rows[0]["username"], "postgres")
+            self.assertEqual(rows[0]["password"], "pw")
+
+    def test_replaces_existing_star_entry(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            run_seed(p, "bigbluebutton", "*", "postgres", "pw1")
+            run_seed(p, "bigbluebutton", "*", "postgres", "pw2")
 
-            # Replace with empty database again
-            run_seed(p, "docker.test", "", "alice", "pw2")
             rows = read_csv_semicolon(p)
-
             self.assertEqual(len(rows), 1)
-            self.assertEqual(rows[0]["database"], "")
+            self.assertEqual(rows[0]["database"], "*")
             self.assertEqual(rows[0]["password"], "pw2")
+
+    def test_rejects_empty_database_value(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            cp = run_seed_expect_fail(p, "docker.test", "", "alice", "pw")
+            self.assertNotEqual(cp.returncode, 0)
+
+            combined = ((cp.stdout or "") + "\n" + (cp.stderr or "")).lower()
+            self.assertIn("error:", combined)
+            self.assertIn("database", combined)
+            self.assertIn("not empty", combined)
+
+            self.assertFalse(p.exists(), "Should not create file on invalid input")
+
+    def test_rejects_invalid_database_name_characters(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            cp = run_seed_expect_fail(p, "docker.test", "app db", "alice", "pw")
+            self.assertNotEqual(cp.returncode, 0)
+
+            combined = ((cp.stdout or "") + "\n" + (cp.stderr or "")).lower()
+            self.assertIn("error:", combined)
+            self.assertIn("invalid database name", combined)
+
+            self.assertFalse(p.exists(), "Should not create file on invalid input")
+
+    def test_rejects_nan_database_name(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            cp = run_seed_expect_fail(p, "docker.test", "nan", "alice", "pw")
+            self.assertNotEqual(cp.returncode, 0)
+
+            combined = ((cp.stdout or "") + "\n" + (cp.stderr or "")).lower()
+            self.assertIn("error:", combined)
+            self.assertIn("must not be 'nan'", combined)
+
+            self.assertFalse(p.exists(), "Should not create file on invalid input")
+
+    def test_accepts_hyphen_and_underscore_database_names(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            run_seed(p, "docker.test", "my_db-1", "alice", "pw")
+
+            rows = read_csv_semicolon(p)
+            self.assertEqual(len(rows), 1)
+            self.assertEqual(rows[0]["database"], "my_db-1")
+
+    def test_file_is_semicolon_delimited_and_has_header(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            p = Path(td) / "databases.csv"
+
+            run_seed(p, "docker.test", "appdb", "alice", "pw")
+
+            txt = read_text(p)
+            self.assertTrue(
+                txt.startswith("instance;database;username;password"),
+                f"Unexpected header / delimiter in file:\n{txt}",
+            )
+            self.assertIn(";", txt)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/__init__.py

@@ -0,0 +1,10 @@
+"""The smallest argv the backup CLI accepts, shared by every test that drives it."""
+
+REQUIRED_PAIRS = [
+    ("--compose-dir", "/compose"),
+    ("--backups-dir", "/backups"),
+    ("--repo-name", "stack"),
+    ("--databases-csv", "/etc/baudolo/databases.csv"),
+]
+REQUIRED = [arg for pair in REQUIRED_PAIRS for arg in pair]
+BASE_ARGV = ["baudolo", *REQUIRED]

tests/unit/backup/compose_fixture.py

@@ -0,0 +1,47 @@
+"""Builds the on-disk compose directories the compose tests discover."""
+
+from __future__ import annotations
+
+import shutil
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from pathlib import Path
+
+
+def touch(p: Path) -> None:
+    p.parent.mkdir(parents=True, exist_ok=True)
+
+    # ".env/env" leaves ".env" behind as a directory, which blocks a later ".env" file.
+    if p.exists() and p.is_dir():
+        shutil.rmtree(p)
+
+    p.write_text("x", encoding="utf-8")
+
+
+def setup_compose_dir(
+    tmp_path: Path,
+    name: str = "mailu",
+    *,
+    compose_name: str = "docker-compose.yml",
+    with_override: bool = False,
+    with_ca_override: bool = False,
+    env_layout: str | None = None,  # None | ".env" | ".env/env"
+) -> Path:
+    d = tmp_path / name
+    d.mkdir(parents=True, exist_ok=True)
+
+    touch(d / compose_name)
+
+    if with_override:
+        touch(d / "docker-compose.override.yml")
+
+    if with_ca_override:
+        touch(d / "docker-compose.ca.override.yml")
+
+    if env_layout == ".env":
+        touch(d / ".env")
+    elif env_layout == ".env/env":
+        touch(d / ".env" / "env")
+
+    return d

tests/unit/backup/test_app_databases_csv.py

@@ -0,0 +1,75 @@
+import io
+import tempfile
+import unittest
+from contextlib import redirect_stderr
+from pathlib import Path
+
+import pandas as pd
+
+# Adjust if your package name/import path differs.
+from baudolo.backup.dumps import load_databases_df
+
+EXPECTED_COLUMNS = ["instance", "database", "username", "password"]
+
+
+class TestLoadDatabasesDf(unittest.TestCase):
+    def test_missing_csv_is_handled_with_warning_and_empty_df(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            missing_path = str(Path(td) / "does-not-exist.csv")
+
+            buf = io.StringIO()
+            with redirect_stderr(buf):
+                df = load_databases_df(missing_path)
+
+            stderr = buf.getvalue()
+            self.assertIn("WARNING:", stderr)
+            self.assertIn("databases.csv not found", stderr)
+
+            self.assertIsInstance(df, pd.DataFrame)
+            self.assertListEqual(list(df.columns), EXPECTED_COLUMNS)
+            self.assertTrue(df.empty)
+
+    def test_empty_csv_is_handled_with_warning_and_empty_df(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            empty_path = Path(td) / "databases.csv"
+            with empty_path.open("w", encoding="utf-8") as f:
+                f.write("")
+
+            buf = io.StringIO()
+            with redirect_stderr(buf):
+                df = load_databases_df(empty_path)
+
+            stderr = buf.getvalue()
+            self.assertIn("WARNING:", stderr)
+            self.assertIn("exists but is empty", stderr)
+
+            self.assertIsInstance(df, pd.DataFrame)
+            self.assertListEqual(list(df.columns), EXPECTED_COLUMNS)
+            self.assertTrue(df.empty)
+
+    def test_valid_csv_loads_without_warning(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            csv_path = Path(td) / "databases.csv"
+
+            content = "instance;database;username;password\nmyapp;*;dbuser;secret\n"
+            with csv_path.open("w", encoding="utf-8") as f:
+                f.write(content)
+
+            buf = io.StringIO()
+            with redirect_stderr(buf):
+                df = load_databases_df(csv_path)
+
+            stderr = buf.getvalue()
+            self.assertEqual(stderr, "")  # no warning expected
+
+            self.assertIsInstance(df, pd.DataFrame)
+            self.assertListEqual(list(df.columns), EXPECTED_COLUMNS)
+            self.assertEqual(len(df), 1)
+            self.assertEqual(df.loc[0, "instance"], "myapp")
+            self.assertEqual(df.loc[0, "database"], "*")
+            self.assertEqual(df.loc[0, "username"], "dbuser")
+            self.assertEqual(df.loc[0, "password"], "secret")
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_app_manifest.py

@@ -0,0 +1,84 @@
+"""What main() records in the manifest for each volume it touched."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup import app
+from baudolo.backup.dumps import VolumeOutcome
+from baudolo.backup.volume import Backing
+
+from . import REQUIRED_PAIRS
+
+ARGV = ["baudolo", *[arg for pair in REQUIRED_PAIRS for arg in pair]]
+
+
+def drive(argv: list[str], dump_result: VolumeOutcome) -> dict:
+    """Run main() over one volume and return the manifest's volume section.
+
+    Args:
+        argv: the command line under test.
+        dump_result: what backup_dumps_for_volume reports.
+    """
+    with (
+        mock.patch("sys.argv", argv),
+        mock.patch.object(app, "get_machine_id", return_value="machine"),
+        mock.patch.object(app, "create_version_directory", return_value="/gen"),
+        mock.patch.object(app, "create_volume_directory", return_value="/gen/vol"),
+        mock.patch.object(app, "load_databases_df", return_value=None),
+        mock.patch.object(app, "docker_volume_names", return_value=["pgdata"]),
+        mock.patch.object(app, "containers_using_volume", return_value=["db"]),
+        mock.patch.object(app, "volume_is_fully_ignored", return_value=False),
+        mock.patch.object(app, "backup_dumps_for_volume", return_value=dump_result),
+        mock.patch.object(app, "inspect_backing", return_value=Backing("/data")),
+        mock.patch.object(app, "write_manifest") as manifest,
+        mock.patch.object(app, "stamp_directory"),
+        mock.patch.object(app, "handle_docker_compose_services"),
+        mock.patch("os.path.isdir", return_value=True),
+        mock.patch.object(app, "backup_volume"),
+        mock.patch.object(app, "filter_stoppable", return_value=[]),
+        mock.patch.object(app, "requires_stop", return_value=False),
+        mock.patch.object(app, "change_containers_status"),
+    ):
+        app.main()
+    return manifest.call_args.args[1]
+
+
+class TestManifest(unittest.TestCase):
+    def test_a_database_volume_without_a_dump_is_recorded_as_undumped(self) -> None:
+        volumes = drive(
+            [*ARGV, "--only-sql"],
+            VolumeOutcome(database=True, dumped=False, engine="postgres"),
+        )
+        self.assertEqual(
+            volumes["pgdata"],
+            VolumeOutcome(database=True, dumped=False, engine="postgres"),
+        )
+
+    def test_a_dumped_database_volume_is_recorded_as_dumped(self) -> None:
+        volumes = drive(
+            [*ARGV, "--only-sql"],
+            VolumeOutcome(database=True, dumped=True, engine="mariadb"),
+        )
+        self.assertEqual(volumes["pgdata"].dumped, True)
+        self.assertEqual(volumes["pgdata"].engine, "mariadb")
+
+    def test_a_plain_volume_is_recorded_as_no_database(self) -> None:
+        volumes = drive(ARGV, VolumeOutcome(database=False, dumped=False))
+        self.assertEqual(volumes["pgdata"].database, False)
+        self.assertIsNone(volumes["pgdata"].engine)
+
+    def test_the_dumped_volume_is_recorded_even_though_the_copy_is_skipped(
+        self,
+    ) -> None:
+        """--only-sql returns to the loop head on success, before the copy."""
+        volumes = drive(
+            [*ARGV, "--only-sql"],
+            VolumeOutcome(database=True, dumped=True, engine="postgres"),
+        )
+        self.assertIn("pgdata", volumes)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_app_only_files.py

@@ -0,0 +1,66 @@
+"""Contract of --only-files: no dump is attempted, every volume is copied."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup import app
+from baudolo.backup.volume import Backing
+
+from . import REQUIRED_PAIRS
+
+ARGV_WITHOUT_CSV = [
+    "baudolo",
+    *[arg for pair in REQUIRED_PAIRS if pair[0] != "--databases-csv" for arg in pair],
+    "--only-files",
+]
+
+
+def drive(argv: list[str]) -> tuple[list[str], list, list]:
+    backed_up: list[str] = []
+
+    def record_backup(versions_dir, volume_name, volume_dir, *, authoritative, source):
+        backed_up.append(volume_name)
+
+    with (
+        mock.patch("sys.argv", argv),
+        mock.patch.object(app, "get_machine_id", return_value="machine"),
+        mock.patch.object(app, "create_version_directory", return_value="/gen"),
+        mock.patch.object(app, "create_volume_directory", return_value="/gen/vol"),
+        mock.patch.object(app, "load_databases_df") as load_csv,
+        mock.patch.object(app, "docker_volume_names", return_value=["pgdata"]),
+        mock.patch.object(app, "containers_using_volume", return_value=["db"]),
+        mock.patch.object(app, "volume_is_fully_ignored", return_value=False),
+        mock.patch.object(app, "backup_dumps_for_volume") as dumps,
+        mock.patch.object(app, "inspect_backing", return_value=Backing("/data")),
+        mock.patch.object(app, "write_manifest"),
+        mock.patch.object(app, "stamp_directory"),
+        mock.patch.object(app, "handle_docker_compose_services"),
+        mock.patch("os.path.isdir", return_value=True),
+        mock.patch.object(app, "backup_volume", side_effect=record_backup),
+        mock.patch.object(app, "filter_stoppable", return_value=[]),
+        mock.patch.object(app, "requires_stop", return_value=False),
+        mock.patch.object(app, "change_containers_status"),
+    ):
+        app.main()
+    return backed_up, dumps.mock_calls, load_csv.mock_calls
+
+
+class TestOnlyFiles(unittest.TestCase):
+    def test_no_dump_is_attempted(self) -> None:
+        _backed_up, dumps, _load_csv = drive(ARGV_WITHOUT_CSV)
+        self.assertEqual(dumps, [])
+
+    def test_the_databases_csv_is_never_read(self) -> None:
+        """It may legitimately be absent, so reading it would abort the run."""
+        _backed_up, _dumps, load_csv = drive(ARGV_WITHOUT_CSV)
+        self.assertEqual(load_csv, [])
+
+    def test_the_volume_is_still_copied(self) -> None:
+        backed_up, _dumps, _load_csv = drive(ARGV_WITHOUT_CSV)
+        self.assertEqual(backed_up, ["pgdata"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_app_snapshot.py

@@ -0,0 +1,92 @@
+"""Contract of app.main's snapshot branch - the caller that runs in production."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup import app
+from baudolo.backup import snapshot as snapshot_mod
+from baudolo.backup.snapshot import volume_snapshot
+from baudolo.backup.volume import Backing
+
+from . import BASE_ARGV
+
+
+def stubbed_snapshot(kind: str, subject: str, tag: str):
+    return volume_snapshot(kind, subject, tag, run=lambda command: [])
+
+
+ARGV = [
+    *BASE_ARGV,
+    "--snapshot",
+    "btrfs",
+    "--snapshot-subject",
+    "/var/lib/docker",
+]
+
+
+def drive(*, present: bool = True, reason: str | None = None) -> list[dict]:
+    calls: list[dict] = []
+
+    def record(versions_dir, volume_name, volume_dir, *, authoritative, source):
+        calls.append(
+            {"volume": volume_name, "authoritative": authoritative, "source": source}
+        )
+
+    with (
+        mock.patch("sys.argv", ARGV),
+        mock.patch.object(app, "get_machine_id", return_value="machine"),
+        mock.patch.object(app, "create_version_directory", return_value="/gen"),
+        mock.patch.object(app, "create_volume_directory", return_value="/gen/vol"),
+        mock.patch.object(app, "load_databases_df", return_value=None),
+        mock.patch.object(app, "docker_volume_names", return_value=["vol"]),
+        mock.patch.object(app, "containers_using_volume", return_value=[]),
+        mock.patch.object(app, "volume_is_fully_ignored", return_value=False),
+        mock.patch.object(app, "backup_dumps_for_volume", return_value=(False, False)),
+        mock.patch.object(
+            app,
+            "inspect_backing",
+            return_value=Backing("/var/lib/docker/volumes/vol/_data"),
+        ),
+        mock.patch.object(snapshot_mod, "unsnapshotted", return_value=reason),
+        mock.patch.object(app, "write_manifest"),
+        mock.patch.object(app, "stamp_directory"),
+        mock.patch.object(app, "handle_docker_compose_services"),
+        mock.patch("os.path.isdir", return_value=present),
+        mock.patch.object(app, "backup_volume", side_effect=record),
+        mock.patch.object(app, "volume_snapshot", stubbed_snapshot),
+    ):
+        app.main()
+    return calls
+
+
+class TestSnapshotBranch(unittest.TestCase):
+    def test_it_passes_a_path_ending_in_a_separator(self) -> None:
+        source = drive(present=True)[0]["source"]
+        self.assertTrue(source.endswith("/volumes/vol/_data/"), source)
+        self.assertNotIn("/var/lib/docker/volumes", source)
+
+    def test_it_reads_from_the_snapshot_and_not_from_the_live_tree(self) -> None:
+        source = drive(present=True)[0]["source"]
+        self.assertTrue(source.startswith("/var/lib/docker/.baudolo-"), source)
+
+    def test_it_compares_by_content_against_the_previous_generation(self) -> None:
+        self.assertTrue(drive(present=True)[0]["authoritative"])
+
+    def test_a_volume_missing_from_the_snapshot_is_copied_live(self) -> None:
+        call = drive(present=False)[0]
+        self.assertEqual(call["source"], "/var/lib/docker/volumes/vol/_data/")
+        self.assertFalse(call["authoritative"])
+
+    def test_a_volume_with_its_own_backing_store_is_copied_live(self) -> None:
+        call = drive(reason="it declares its own backing store")[0]
+        self.assertEqual(call["source"], "/var/lib/docker/volumes/vol/_data/")
+        self.assertFalse(
+            call["authoritative"],
+            "the snapshot holds an empty directory for it, not its data",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_app_volumes_no_backup_required.py

@@ -0,0 +1,82 @@
+"""Contract of --volumes-no-backup-required: exclusion is per volume name,
+independent of which containers use it."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup import app
+from baudolo.backup.volume import Backing
+
+from . import BASE_ARGV
+
+ARGV = [
+    *BASE_ARGV,
+    "--volumes-no-backup-required",
+    "derived",
+]
+
+
+def drive() -> tuple[list[str], list[str], list[str]]:
+    backed_up: list[str] = []
+    created: list[str] = []
+    inspected: list[str] = []
+
+    def record_backup(versions_dir, volume_name, volume_dir, *, authoritative, source):
+        backed_up.append(volume_name)
+
+    with (
+        mock.patch("sys.argv", ARGV),
+        mock.patch.object(app, "get_machine_id", return_value="machine"),
+        mock.patch.object(app, "create_version_directory", return_value="/gen"),
+        mock.patch.object(
+            app,
+            "create_volume_directory",
+            side_effect=lambda _version_dir, name: created.append(name) or "/gen/vol",
+        ),
+        mock.patch.object(app, "load_databases_df", return_value=None),
+        mock.patch.object(
+            app, "docker_volume_names", return_value=["derived", "state"]
+        ),
+        mock.patch.object(
+            app,
+            "containers_using_volume",
+            side_effect=lambda name: inspected.append(name) or ["app"],
+        ),
+        mock.patch.object(app, "volume_is_fully_ignored", return_value=False),
+        mock.patch.object(app, "backup_dumps_for_volume", return_value=(False, False)),
+        mock.patch.object(app, "inspect_backing", return_value=Backing("/data")),
+        mock.patch.object(app, "write_manifest"),
+        mock.patch.object(app, "stamp_directory"),
+        mock.patch.object(app, "handle_docker_compose_services"),
+        mock.patch("os.path.isdir", return_value=True),
+        mock.patch.object(app, "backup_volume", side_effect=record_backup),
+        mock.patch.object(app, "filter_stoppable", return_value=[]),
+        mock.patch.object(app, "requires_stop", return_value=False),
+        mock.patch.object(app, "change_containers_status"),
+    ):
+        app.main()
+    return backed_up, created, inspected
+
+
+class TestVolumesNoBackupRequired(unittest.TestCase):
+    def test_the_named_volume_is_never_backed_up(self) -> None:
+        backed_up, _created, _inspected = drive()
+        self.assertNotIn("derived", backed_up)
+
+    def test_a_sibling_volume_of_the_same_container_survives(self) -> None:
+        backed_up, _created, _inspected = drive()
+        self.assertEqual(backed_up, ["state"])
+
+    def test_no_generation_directory_is_created_for_it(self) -> None:
+        _backed_up, created, _inspected = drive()
+        self.assertEqual(created, ["state"])
+
+    def test_the_skip_precedes_the_container_inspection(self) -> None:
+        _backed_up, _created, inspected = drive()
+        self.assertEqual(inspected, ["state"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_cli.py

@@ -0,0 +1,110 @@
+"""Contract of the backup CLI, in particular the snapshot flag pairing."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup.cli import parse_args
+
+from . import REQUIRED, REQUIRED_PAIRS
+
+
+def parse(*extra: str):
+    with mock.patch("sys.argv", ["baudolo", *REQUIRED, *extra]):
+        return parse_args()
+
+
+class TestSnapshotFlags(unittest.TestCase):
+    def test_no_snapshot_by_default(self) -> None:
+        args = parse()
+        self.assertIsNone(args.snapshot)
+        self.assertIsNone(args.snapshot_subject)
+
+    def test_both_flags_together_are_accepted(self) -> None:
+        args = parse("--snapshot", "btrfs", "--snapshot-subject", "/var/lib/docker")
+        self.assertEqual(args.snapshot, "btrfs")
+        self.assertEqual(args.snapshot_subject, "/var/lib/docker")
+
+    def test_the_kind_alone_is_rejected(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse("--snapshot", "btrfs")
+
+    def test_the_subject_alone_is_rejected(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse("--snapshot-subject", "/var/lib/docker")
+
+    def test_an_unsupported_kind_is_rejected(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse("--snapshot", "ext4", "--snapshot-subject", "/var/lib/docker")
+
+    def test_zfs_is_accepted(self) -> None:
+        self.assertEqual(
+            parse("--snapshot", "zfs", "--snapshot-subject", "/d").snapshot, "zfs"
+        )
+
+    def test_shutdown_is_rejected_because_nothing_is_stopped(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse("--snapshot", "btrfs", "--snapshot-subject", "/d", "--shutdown")
+
+    def test_shutdown_stays_available_without_a_snapshot(self) -> None:
+        self.assertTrue(parse("--shutdown").shutdown)
+
+    def test_hard_restart_is_rejected_because_nothing_is_stopped(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse(
+                "--snapshot",
+                "btrfs",
+                "--snapshot-subject",
+                "/d",
+                "--hard-restart-projects",
+                "mailu",
+            )
+
+    def test_hard_restart_stays_available_without_a_snapshot(self) -> None:
+        self.assertEqual(
+            parse("--hard-restart-projects", "mailu").hard_restart_projects, ["mailu"]
+        )
+
+
+class TestRequiredFlags(unittest.TestCase):
+    def test_no_flag_falls_back_to_a_default(self) -> None:
+        for omitted, _ in REQUIRED_PAIRS:
+            argv = [a for pair in REQUIRED_PAIRS if pair[0] != omitted for a in pair]
+            with (
+                self.subTest(omitted=omitted),
+                mock.patch("sys.argv", ["baudolo", *argv]),
+                self.assertRaises(SystemExit),
+            ):
+                parse_args()
+
+
+class TestBackupScope(unittest.TestCase):
+    """--only-sql and --only-files name the two halves a generation can hold."""
+
+    def test_both_halves_by_default(self) -> None:
+        args = parse()
+        self.assertFalse(args.only_sql)
+        self.assertFalse(args.only_files)
+
+    def test_either_half_alone_is_accepted(self) -> None:
+        self.assertTrue(parse("--only-sql").only_sql)
+        self.assertTrue(parse("--only-files").only_files)
+
+    def test_asking_for_both_halves_alone_is_rejected(self) -> None:
+        with self.assertRaises(SystemExit):
+            parse("--only-sql", "--only-files")
+
+    def test_only_files_needs_no_databases_csv(self) -> None:
+        argv = [
+            arg
+            for pair in REQUIRED_PAIRS
+            if pair[0] != "--databases-csv"
+            for arg in pair
+        ]
+        with mock.patch("sys.argv", ["baudolo", *argv, "--only-files"]):
+            self.assertIsNone(parse_args().databases_csv)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_compose.py

@@ -0,0 +1,212 @@
+from __future__ import annotations
+
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+from .compose_fixture import setup_compose_dir as _setup_compose_dir
+
+
+class TestCompose(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls) -> None:
+        from baudolo.backup import compose as mod
+
+        cls.compose_mod = mod
+
+    def test_find_compose_file_supports_all_valid_names_case_insensitive(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+
+            variants = [
+                "compose.yml",
+                "compose.yaml",
+                "docker-compose.yml",
+                "docker-compose.yaml",
+                "docker-compose.yAml",
+            ]
+
+            for i, name in enumerate(variants):
+                d = _setup_compose_dir(
+                    tmp_path,
+                    name=f"project{i}",
+                    compose_name=name,
+                )
+                found = self.compose_mod._find_compose_file(str(d))
+                self.assertIsNotNone(found)
+                self.assertEqual(found.name, name)
+
+    def test_find_compose_file_returns_none_when_missing(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+            d = tmp_path / "empty"
+            d.mkdir(parents=True, exist_ok=True)
+
+            found = self.compose_mod._find_compose_file(str(d))
+            self.assertIsNone(found)
+
+    def test_build_cmd_uses_wrapper_when_present(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+            d = _setup_compose_dir(
+                tmp_path,
+                with_override=True,
+                with_ca_override=True,
+                env_layout=".env",
+            )
+
+            def fake_which(name: str):
+                if name == "compose":
+                    return "/usr/local/bin/compose"
+                return None
+
+            with patch.object(self.compose_mod.shutil, "which", fake_which):
+                cmd = self.compose_mod._build_compose_cmd(str(d), ["up", "-d"])
+
+            self.assertEqual(
+                cmd,
+                [
+                    "/usr/local/bin/compose",
+                    "--chdir",
+                    str(d.resolve()),
+                    "--",
+                    "up",
+                    "-d",
+                ],
+            )
+
+    def test_build_cmd_fallback_uses_plain_docker_compose_chdir(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+            d = _setup_compose_dir(
+                tmp_path,
+                with_override=True,
+                with_ca_override=True,
+                env_layout=".env",
+            )
+
+            def fake_which(name: str):
+                if name == "compose":
+                    return None
+                if name == "docker":
+                    return "/usr/bin/docker"
+                return None
+
+            with patch.object(self.compose_mod.shutil, "which", fake_which):
+                cmd = self.compose_mod._build_compose_cmd(
+                    str(d), ["up", "-d", "--force-recreate"]
+                )
+
+            expected: list[str] = [
+                "/usr/bin/docker",
+                "compose",
+                "--chdir",
+                str(d.resolve()),
+                "up",
+                "-d",
+                "--force-recreate",
+            ]
+            self.assertEqual(cmd, expected)
+
+    def test_hard_restart_calls_run_twice_with_correct_cmds_wrapper(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+            d = _setup_compose_dir(tmp_path, name="mailu", env_layout=".env")
+
+            def fake_which(name: str):
+                if name == "compose":
+                    return "/usr/local/bin/compose"
+                return None
+
+            with patch.object(self.compose_mod.shutil, "which", fake_which):
+                calls = []
+
+                def fake_run(cmd, check: bool):
+                    calls.append((cmd, check))
+                    return 0
+
+                with patch.object(self.compose_mod.subprocess, "run", fake_run):
+                    self.compose_mod.hard_restart_docker_services(str(d))
+
+            self.assertEqual(
+                calls,
+                [
+                    (
+                        [
+                            "/usr/local/bin/compose",
+                            "--chdir",
+                            str(d.resolve()),
+                            "--",
+                            "down",
+                        ],
+                        True,
+                    ),
+                    (
+                        [
+                            "/usr/local/bin/compose",
+                            "--chdir",
+                            str(d.resolve()),
+                            "--",
+                            "up",
+                            "-d",
+                        ],
+                        True,
+                    ),
+                ],
+            )
+
+    def test_hard_restart_calls_run_twice_with_correct_cmds_fallback(self) -> None:
+        with tempfile.TemporaryDirectory() as td:
+            tmp_path = Path(td)
+            d = _setup_compose_dir(
+                tmp_path,
+                name="mailu",
+                with_override=True,
+                with_ca_override=True,
+                env_layout=".env/env",
+            )
+
+            def fake_which(name: str):
+                if name == "compose":
+                    return None
+                if name == "docker":
+                    return "/usr/bin/docker"
+                return None
+
+            with patch.object(self.compose_mod.shutil, "which", fake_which):
+                calls = []
+
+                def fake_run(cmd, check: bool):
+                    calls.append((cmd, check))
+                    return 0
+
+                with patch.object(self.compose_mod.subprocess, "run", fake_run):
+                    self.compose_mod.hard_restart_docker_services(str(d))
+
+            self.assertEqual(
+                calls,
+                [
+                    (
+                        [
+                            "/usr/bin/docker",
+                            "compose",
+                            "--chdir",
+                            str(d.resolve()),
+                            "down",
+                        ],
+                        True,
+                    ),
+                    (
+                        [
+                            "/usr/bin/docker",
+                            "compose",
+                            "--chdir",
+                            str(d.resolve()),
+                            "up",
+                            "-d",
+                        ],
+                        True,
+                    ),
+                ],
+            )

tests/unit/backup/test_compose_hard_restart.py

@@ -0,0 +1,45 @@
+from __future__ import annotations
+
+import unittest
+from unittest.mock import patch
+
+from . import BASE_ARGV
+
+
+class HardRestartArgTests(unittest.TestCase):
+    """The hard-restart list defaults to empty (no compose down/up); callers
+    opt in per dir, e.g. compose hosts pass 'mailu' while swarm hosts, where
+    the dir is a stack whose overlay network collides with compose up, pass
+    nothing."""
+
+    def _parse(self, extra: list[str]):
+        import sys
+
+        from baudolo.backup import cli
+
+        argv = [
+            *BASE_ARGV,
+            "--database-containers",
+            "postgres",
+            "--images-no-stop-required",
+            "redis",
+            *extra,
+        ]
+        with patch.object(sys, "argv", argv):
+            return cli.parse_args()
+
+    def test_default_is_empty(self) -> None:
+        args = self._parse([])
+        self.assertEqual(args.hard_restart_projects, [])
+
+    def test_empty_flag_stays_empty(self) -> None:
+        args = self._parse(["--hard-restart-projects"])
+        self.assertEqual(args.hard_restart_projects, [])
+
+    def test_explicit_names_preserved(self) -> None:
+        args = self._parse(["--hard-restart-projects", "mailu", "foo"])
+        self.assertEqual(args.hard_restart_projects, ["mailu", "foo"])
+
+
+if __name__ == "__main__":
+    unittest.main(verbosity=2)

tests/unit/backup/test_db_mariadb_dump.py

@@ -0,0 +1,103 @@
+import tempfile
+import unittest
+from unittest.mock import patch
+
+import pandas as pd
+
+from baudolo.backup import db as db_mod
+
+
+def _df(rows):
+    return pd.DataFrame(rows, columns=["instance", "database", "username", "password"])
+
+
+def _capture_dumps(*, db_type, rows, container, dump_tool="mariadb-dump"):
+    """Every (argv, env) the dump path would have run."""
+    captured = []
+
+    def _capture(command, out_file, *, env=None):
+        captured.append((list(command), env))
+
+    with (
+        tempfile.TemporaryDirectory() as td,
+        patch.object(db_mod, "execute_to_file", side_effect=_capture),
+    ):
+        db_mod.backup_database(
+            container=container,
+            volume_dir=td,
+            db_type=db_type,
+            dump_tool=dump_tool,
+            databases_df=_df(rows),
+            database_containers=[container],
+        )
+    return captured
+
+
+class TestMariaDBDumpUsesTCP(unittest.TestCase):
+    # Regression guard for 'Access denied for user <user>@localhost' when only
+    # '<user>'@'%' is granted: the in-container mariadb-dump MUST force TCP so
+    # the connection is auth-matched against '%' instead of socket->localhost.
+
+    def test_mariadb_dump_forces_tcp_loopback(self):
+        captured = _capture_dumps(
+            db_type="mariadb",
+            rows=[("mariadb", "appdb", "appuser", "s3cret")],
+            container="mariadb",
+        )
+        self.assertEqual(len(captured), 1, f"expected one dump, got: {captured}")
+
+        argv, env = captured[0]
+        self.assertEqual(argv[:3], ["docker", "exec", "mariadb"])
+        self.assertIn("--protocol=tcp", argv)
+        self.assertEqual(argv[argv.index("-h") + 1], "127.0.0.1")
+        self.assertEqual(argv[argv.index("-u") + 1], "appuser")
+        self.assertIn("-ps3cret", argv)
+        self.assertEqual(argv[-1], "appdb")
+        self.assertIsNone(env)
+
+    def test_the_probed_client_is_the_one_invoked(self):
+        captured = _capture_dumps(
+            db_type="mariadb",
+            rows=[("mariadb", "appdb", "appuser", "s3cret")],
+            container="mariadb",
+            dump_tool="mysqldump",
+        )
+        argv, _env = captured[0]
+        self.assertIn("mysqldump", argv)
+        self.assertNotIn("mariadb-dump", argv)
+
+    def test_postgres_dump_unaffected(self):
+        captured = _capture_dumps(
+            db_type="postgres",
+            rows=[("pg", "appdb", "appuser", "s3cret")],
+            container="pg",
+        )
+        argv, _env = captured[0]
+        self.assertIn("pg_dump", argv)
+        self.assertNotIn("--protocol=tcp", argv)
+
+    def test_the_password_travels_in_the_environment_not_the_argv(self):
+        """A process listing shows argv; PGPASSWORD must not be in it."""
+        captured = _capture_dumps(
+            db_type="postgres",
+            rows=[("pg", "appdb", "appuser", "s3cret")],
+            container="pg",
+        )
+        argv, env = captured[0]
+        self.assertEqual(env, {"PGPASSWORD": "s3cret"})
+        self.assertNotIn("s3cret", argv)
+
+
+class TestNoShellReachesTheDump(unittest.TestCase):
+    def test_a_hostile_database_name_never_reaches_a_command(self):
+        """validate_database refuses it, so no argv is built at all."""
+        with self.assertRaises(ValueError):
+            _capture_dumps(
+                db_type="postgres",
+                rows=[("pg", "app;rm -rf /", "appuser", "s3cret")],
+                container="pg",
+            )
+
+
+if __name__ == "__main__":
+    unittest.main(verbosity=2)

tests/unit/backup/test_docker_exec_env.py

@@ -0,0 +1,61 @@
+"""How a secret reaches the command running inside the container."""
+
+from __future__ import annotations
+
+import unittest
+
+from baudolo.backup.db import fallback_pg_dumpall
+from baudolo.backup.docker import docker_exec_argv
+
+
+class TestForwardEnv(unittest.TestCase):
+    def test_nothing_is_added_when_no_variable_is_named(self) -> None:
+        self.assertEqual(
+            docker_exec_argv("c1", ["true"]),
+            ["docker", "exec", "c1", "true"],
+        )
+
+    def test_a_named_variable_is_forwarded_without_its_value(self) -> None:
+        """-e NAME=value would publish the secret in the host's process list."""
+        argv = docker_exec_argv("c1", ["true"], forward_env=["PGPASSWORD"])
+        self.assertEqual(argv, ["docker", "exec", "-e", "PGPASSWORD", "c1", "true"])
+
+    def test_the_flag_precedes_the_container(self) -> None:
+        """docker reads options before the container name, arguments after it."""
+        argv = docker_exec_argv(
+            "c1", ["pg_dump", "-U", "u"], interactive=True, forward_env=["PGPASSWORD"]
+        )
+        self.assertLess(argv.index("-e"), argv.index("c1"))
+        self.assertLess(argv.index("-i"), argv.index("c1"))
+        self.assertGreater(argv.index("pg_dump"), argv.index("c1"))
+
+    def test_several_variables_each_get_their_own_flag(self) -> None:
+        argv = docker_exec_argv("c1", ["true"], forward_env=["A", "B"])
+        self.assertEqual(argv[:6], ["docker", "exec", "-e", "A", "-e", "B"])
+
+
+class TestPostgresDumpCarriesThePassword(unittest.TestCase):
+    def test_the_cluster_dump_forwards_pgpassword(self) -> None:
+        seen: dict = {}
+
+        def fake(command, out_file, *, env=None):
+            seen["command"] = command
+            seen["env"] = env
+
+        import baudolo.backup.db as db
+
+        original = db.execute_to_file
+        db.execute_to_file = fake
+        try:
+            fallback_pg_dumpall("pg", "user", "secret", "/tmp/out.sql")
+        finally:
+            db.execute_to_file = original
+
+        self.assertIn("-e", seen["command"])
+        self.assertEqual(seen["command"][seen["command"].index("-e") + 1], "PGPASSWORD")
+        self.assertEqual(seen["env"], {"PGPASSWORD": "secret"})
+        self.assertNotIn("secret", seen["command"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_docker_swarm.py

@@ -0,0 +1,53 @@
+import unittest
+from unittest.mock import patch
+
+from baudolo.backup import docker as docker_mod
+from baudolo.backup.shell import BackupError
+
+
+class TestIsSwarmTask(unittest.TestCase):
+    @patch.object(docker_mod, "execute_shell_command", return_value=["task-id-123"])
+    def test_true_when_task_label_present(self, _mock) -> None:
+        self.assertTrue(docker_mod.is_swarm_task("c1"))
+
+    @patch.object(docker_mod, "execute_shell_command", return_value=[""])
+    def test_false_when_label_empty(self, _mock) -> None:
+        self.assertFalse(docker_mod.is_swarm_task("c1"))
+
+    @patch.object(docker_mod, "execute_shell_command", return_value=[])
+    def test_false_when_no_output(self, _mock) -> None:
+        self.assertFalse(docker_mod.is_swarm_task("c1"))
+
+    @patch.object(
+        docker_mod,
+        "execute_shell_command",
+        side_effect=[BackupError("gone"), []],
+    )
+    def test_vanished_container_counts_as_not_stoppable(self, _mock) -> None:
+        # A container removed between listing and inspect must not abort the
+        # whole backup run; treating it as a swarm task keeps it out of every
+        # stop/start and image-inspect path.
+        self.assertTrue(docker_mod.is_swarm_task("gone-container"))
+
+    @patch.object(
+        docker_mod,
+        "execute_shell_command",
+        side_effect=[BackupError("daemon hiccup"), ["still-here"]],
+    )
+    def test_inspect_failure_on_existing_container_still_fails(self, _mock) -> None:
+        # If the container still exists, an inspect failure must keep failing
+        # the run: silently skipping the stop would back up a hot volume and
+        # report green without the stop guarantee.
+        with self.assertRaises(BackupError):
+            docker_mod.is_swarm_task("still-here")
+
+
+class TestFilterStoppable(unittest.TestCase):
+    @patch.object(docker_mod, "is_swarm_task", side_effect=[False, True, False])
+    def test_mixed_list_keeps_order_and_drops_tasks(self, _mock) -> None:
+        result = docker_mod.filter_stoppable(["plain-1", "swarm-task", "plain-2"])
+        self.assertEqual(result, ["plain-1", "plain-2"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_docker_tool_probe.py

@@ -0,0 +1,44 @@
+import unittest
+from unittest.mock import patch
+
+from baudolo.backup import docker as docker_mod
+from baudolo.backup.shell import BackupError
+
+
+class TestImageId(unittest.TestCase):
+    def test_the_id_is_returned_without_surrounding_whitespace(self) -> None:
+        with patch.object(
+            docker_mod, "execute_shell_command", return_value=["sha256:abc \n"]
+        ):
+            self.assertEqual(docker_mod.image_id("c1"), "sha256:abc")
+
+
+class TestHasTool(unittest.TestCase):
+    def test_a_tool_that_runs_is_present(self) -> None:
+        with patch.object(docker_mod, "execute_shell_command", return_value=[]):
+            self.assertTrue(docker_mod.has_tool("c1", "pg_dumpall"))
+
+    def test_a_tool_that_exits_non_zero_is_absent(self) -> None:
+        with patch.object(
+            docker_mod, "execute_shell_command", side_effect=BackupError("127")
+        ):
+            self.assertFalse(docker_mod.has_tool("c1", "mariadb-dump"))
+
+    def test_the_probe_needs_no_shell_in_the_image(self) -> None:
+        """A distroless database ships no shell; `sh -c` would deny every tool."""
+        captured = []
+
+        def _capture(cmd):
+            captured.append(cmd)
+            return []
+
+        with patch.object(docker_mod, "execute_shell_command", side_effect=_capture):
+            docker_mod.has_tool("c1", "pg_dumpall")
+
+        self.assertEqual(
+            captured, [["docker", "exec", "c1", "pg_dumpall", "--version"]]
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_dumps_engine_detection.py

@@ -0,0 +1,137 @@
+import unittest
+from unittest.mock import patch
+
+import pandas as pd
+
+from baudolo.backup import dumps as dumps_mod
+
+
+def _df(rows):
+    return pd.DataFrame(rows, columns=["instance", "database", "username", "password"])
+
+
+class _Probe:
+    def __init__(self, available, image="sha256:aaa"):
+        self.available = set(available)
+        self.image = image
+        self.calls = []
+
+    def has_tool(self, container, tool):
+        self.calls.append((container, tool))
+        return tool in self.available
+
+    def image_id(self, container):
+        return self.image if isinstance(self.image, str) else self.image[container]
+
+
+def _detect(probe, container="c1"):
+    dumps_mod._ENGINE_BY_IMAGE.clear()
+    with (
+        patch.object(dumps_mod, "has_tool", probe.has_tool),
+        patch.object(dumps_mod, "image_id", probe.image_id),
+    ):
+        return dumps_mod.container_engine(container)
+
+
+class TestContainerEngine(unittest.TestCase):
+    def test_a_postgres_is_found_by_its_dump_tool(self):
+        self.assertEqual(_detect(_Probe(["pg_dumpall"])), ("postgres", "pg_dumpall"))
+
+    def test_a_mariadb_is_found_by_its_dump_tool(self):
+        self.assertEqual(_detect(_Probe(["mariadb-dump"])), ("mariadb", "mariadb-dump"))
+
+    def test_an_image_with_only_mysqldump_is_dumped_with_mysqldump(self):
+        self.assertEqual(_detect(_Probe(["mysqldump"])), ("mariadb", "mysqldump"))
+
+    def test_a_container_without_either_tool_is_no_database(self):
+        self.assertIsNone(_detect(_Probe([])))
+
+    def test_the_probe_stops_at_the_first_tool_it_finds(self):
+        probe = _Probe(["pg_dumpall", "mariadb-dump"])
+        _detect(probe)
+        self.assertEqual(probe.calls, [("c1", "pg_dumpall")])
+
+    def test_the_image_name_does_not_decide_the_engine(self):
+        """The trap the old substring test fell into, from both directions."""
+        probe = _Probe(["pg_dumpall"], image="svc-db-mariadb-mgr-01:5000/pg_custom")
+        self.assertEqual(_detect(probe), ("postgres", "pg_dumpall"))
+
+        probe = _Probe(["mariadb-dump"], image="discourse-database:17")
+        self.assertEqual(_detect(probe), ("mariadb", "mariadb-dump"))
+
+
+class TestProbeCache(unittest.TestCase):
+    def test_replicas_of_one_image_are_probed_once(self):
+        probe = _Probe(["pg_dumpall"])
+        dumps_mod._ENGINE_BY_IMAGE.clear()
+        with (
+            patch.object(dumps_mod, "has_tool", probe.has_tool),
+            patch.object(dumps_mod, "image_id", probe.image_id),
+        ):
+            first = dumps_mod.container_engine("replica-1")
+            second = dumps_mod.container_engine("replica-2")
+        self.assertEqual(first, second)
+        self.assertEqual(len(probe.calls), 1)
+
+    def test_a_second_image_is_probed_separately(self):
+        probe = _Probe(["pg_dumpall"], image={"pg": "sha256:aaa", "app": "sha256:bbb"})
+        dumps_mod._ENGINE_BY_IMAGE.clear()
+        with (
+            patch.object(dumps_mod, "has_tool", probe.has_tool),
+            patch.object(dumps_mod, "image_id", probe.image_id),
+        ):
+            self.assertEqual(
+                dumps_mod.container_engine("pg"), ("postgres", "pg_dumpall")
+            )
+            probe.available = set()
+            self.assertIsNone(dumps_mod.container_engine("app"))
+
+
+class TestBackupDispatch(unittest.TestCase):
+    def test_the_probed_tool_reaches_the_dump(self):
+        probe = _Probe(["mysqldump"])
+        seen = {}
+
+        def _fake_backup_database(**kwargs):
+            seen.update(kwargs)
+            return True
+
+        dumps_mod._ENGINE_BY_IMAGE.clear()
+        with (
+            patch.object(dumps_mod, "has_tool", probe.has_tool),
+            patch.object(dumps_mod, "image_id", probe.image_id),
+            patch.object(dumps_mod, "backup_database", _fake_backup_database),
+        ):
+            outcome = dumps_mod.backup_mariadb_or_postgres(
+                container="c1",
+                volume_dir="/tmp",
+                databases_df=_df([("c1", "appdb", "u", "p")]),
+                database_containers=["c1"],
+            )
+
+        self.assertTrue(outcome.database)
+        self.assertTrue(outcome.dumped)
+        self.assertEqual(outcome.engine, "mariadb")
+        self.assertEqual(seen["db_type"], "mariadb")
+        self.assertEqual(seen["dump_tool"], "mysqldump")
+
+    def test_a_non_database_container_is_left_to_the_file_backup(self):
+        probe = _Probe([])
+        dumps_mod._ENGINE_BY_IMAGE.clear()
+        with (
+            patch.object(dumps_mod, "has_tool", probe.has_tool),
+            patch.object(dumps_mod, "image_id", probe.image_id),
+        ):
+            self.assertEqual(
+                dumps_mod.backup_mariadb_or_postgres(
+                    container="c1",
+                    volume_dir="/tmp",
+                    databases_df=_df([]),
+                    database_containers=[],
+                ),
+                dumps_mod.VolumeOutcome(database=False, dumped=False, engine=None),
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_get_instance.py

@@ -0,0 +1,67 @@
+"""Which databases.csv instance a container name resolves to.
+
+The cases are the container names real deployments produce, in both compose
+and swarm, so a change to the regex has to state which shape it gives up.
+"""
+
+from __future__ import annotations
+
+import unittest
+
+from baudolo.backup.db import get_instance
+
+
+class TestDeclaredContainers(unittest.TestCase):
+    def test_a_declared_container_is_its_own_instance(self) -> None:
+        self.assertEqual(
+            get_instance("postgres-central", ["postgres-central"]), "postgres-central"
+        )
+
+    def test_a_declaration_beats_the_regex(self) -> None:
+        """A declared name is taken whole even when it carries a token the
+        fallback would otherwise strip."""
+        self.assertEqual(
+            get_instance("shop-database", ["shop-database"]), "shop-database"
+        )
+
+    def test_an_undeclared_central_engine_resolves_to_nothing(self) -> None:
+        """`postgres-central` has no separator before its token, so nothing is
+        stripped - a central engine has to be declared to be found."""
+        self.assertIsNone(get_instance("postgres-central", []))
+
+
+class TestDedicatedEngines(unittest.TestCase):
+    def test_compose_names_the_container_with_a_hyphen(self) -> None:
+        self.assertEqual(get_instance("discourse-database", []), "discourse")
+
+    def test_swarm_names_the_task_with_an_underscore_and_a_slot(self) -> None:
+        """Swarm suppresses container_name and names the task
+        <stack>_<service>.<slot>.<id>, which must land on the same instance as
+        the compose name so one databases.csv serves both modes."""
+        self.assertEqual(get_instance("discourse_database.1.k3f9x2", []), "discourse")
+
+    def test_an_explicitly_named_engine_keeps_its_entity(self) -> None:
+        self.assertEqual(get_instance("bigbluebutton-postgres-1", []), "bigbluebutton")
+
+    def test_the_short_token_is_stripped_too(self) -> None:
+        self.assertEqual(get_instance("matomo-db", []), "matomo")
+
+    def test_mariadb_uses_the_same_suffix(self) -> None:
+        self.assertEqual(get_instance("matomo-database", []), "matomo")
+
+
+class TestApplicationContainers(unittest.TestCase):
+    def test_a_bare_application_name_is_not_a_database(self) -> None:
+        """Returning the name unchanged here would offer the application as a
+        second engine for its own dedicated database's instance."""
+        self.assertIsNone(get_instance("discourse", []))
+
+    def test_a_swarm_application_task_is_not_a_database(self) -> None:
+        self.assertIsNone(get_instance("discourse_discourse.1.k3f9x2", []))
+
+    def test_an_application_that_merely_starts_with_a_token_is_not_split(self) -> None:
+        self.assertIsNone(get_instance("dbeaver", []))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_layout.py

@@ -0,0 +1,53 @@
+"""Contract of where a backup run puts its directories."""
+
+from __future__ import annotations
+
+import tempfile
+import unittest
+from pathlib import Path
+from unittest import mock
+
+from baudolo.backup import layout as mod
+from baudolo.backup.shell import BackupError
+
+
+class TestVersionDirectory(unittest.TestCase):
+    def test_it_creates_the_generation_directory(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            created = mod.create_version_directory(tmp, "20260731020304")
+            self.assertTrue(Path(created).is_dir())
+            self.assertEqual(Path(created).name, "20260731020304")
+
+    def test_it_refuses_a_generation_another_run_already_claimed(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            mod.create_version_directory(tmp, "20260731")
+            with self.assertRaises(BackupError) as caught:
+                mod.create_version_directory(tmp, "20260731")
+            self.assertIn("20260731", str(caught.exception))
+
+    def test_it_creates_missing_parents(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            nested = str(Path(tmp) / "machine" / "repo")
+            created = mod.create_version_directory(nested, "20260731")
+            self.assertTrue(Path(created).is_dir())
+
+
+class TestVolumeDirectory(unittest.TestCase):
+    def test_it_nests_the_volume_under_the_generation(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            created = mod.create_volume_directory(tmp, "postgres_data")
+            self.assertEqual(Path(created).parent, Path(tmp))
+            self.assertTrue(Path(created).is_dir())
+
+
+class TestMachineId(unittest.TestCase):
+    def test_it_takes_the_hash_without_the_filename(self) -> None:
+        digest = "a" * 64
+        with mock.patch.object(
+            mod, "execute_shell_command", return_value=[f"{digest}  /etc/machine-id"]
+        ):
+            self.assertEqual(mod.get_machine_id(), digest)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_policy.py

@@ -0,0 +1,68 @@
+"""Contract of the rules deciding what is backed up and what must stop."""
+
+from __future__ import annotations
+
+import unittest
+from unittest import mock
+
+from baudolo.backup import policy as mod
+
+
+class TestIsImageIgnored(unittest.TestCase):
+    def test_an_empty_whitelist_ignores_nothing(self) -> None:
+        self.assertFalse(mod.is_image_ignored("c1", []))
+
+    def test_a_listed_image_is_ignored(self) -> None:
+        with mock.patch.object(mod, "get_image_info", return_value="alpine:3.20"):
+            self.assertTrue(mod.is_image_ignored("c1", ["alpine:3.20"]))
+
+    def test_matching_is_exact(self) -> None:
+        with mock.patch.object(mod, "get_image_info", return_value="alpine:3.21"):
+            self.assertFalse(mod.is_image_ignored("c1", ["alpine:3.20"]))
+
+
+class TestVolumeIsFullyIgnored(unittest.TestCase):
+    def test_a_volume_without_containers_is_kept(self) -> None:
+        self.assertFalse(mod.volume_is_fully_ignored([], ["alpine:3.20"]))
+
+    def test_it_needs_every_container_to_be_ignored(self) -> None:
+        with mock.patch.object(mod, "get_image_info", side_effect=["a", "b"]):
+            self.assertFalse(mod.volume_is_fully_ignored(["c1", "c2"], ["a"]))
+
+    def test_all_ignored_skips_the_volume(self) -> None:
+        with mock.patch.object(mod, "get_image_info", side_effect=["a", "a"]):
+            self.assertTrue(mod.volume_is_fully_ignored(["c1", "c2"], ["a"]))
+
+
+class TestRequiresStop(unittest.TestCase):
+    def test_no_containers_means_no_stop(self) -> None:
+        self.assertFalse(mod.requires_stop([], []))
+
+    def test_a_swarm_task_never_forces_a_stop(self) -> None:
+        with mock.patch.object(mod, "is_swarm_task", return_value=True):
+            self.assertFalse(mod.requires_stop(["c1"], []))
+
+    def test_a_whitelisted_image_does_not_force_a_stop(self) -> None:
+        with (
+            mock.patch.object(mod, "is_swarm_task", return_value=False),
+            mock.patch.object(mod, "get_image_info", return_value="alpine:3.20"),
+        ):
+            self.assertFalse(mod.requires_stop(["c1"], ["alpine:3.20"]))
+
+    def test_an_unlisted_image_forces_a_stop(self) -> None:
+        with (
+            mock.patch.object(mod, "is_swarm_task", return_value=False),
+            mock.patch.object(mod, "get_image_info", return_value="postgres:17"),
+        ):
+            self.assertTrue(mod.requires_stop(["c1"], ["alpine:3.20"]))
+
+    def test_one_unlisted_container_is_enough(self) -> None:
+        with (
+            mock.patch.object(mod, "is_swarm_task", return_value=False),
+            mock.patch.object(mod, "get_image_info", side_effect=["alpine:3.20", "x"]),
+        ):
+            self.assertTrue(mod.requires_stop(["c1", "c2"], ["alpine:3.20"]))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_snapshot.py

@@ -0,0 +1,164 @@
+"""Contract of the filesystem snapshot used to capture volumes atomically."""
+
+from __future__ import annotations
+
+import unittest
+
+from baudolo.backup.shell import BackupError
+from baudolo.backup.snapshot import SnapshotError, volume_snapshot
+
+
+class Runner:
+    def __init__(self, replies: dict[str, list[str]] | None = None) -> None:
+        self.calls: list[list[str]] = []
+        self.replies = replies or {}
+
+    def __call__(self, command: list[str]) -> list[str]:
+        self.calls.append(list(command))
+        for prefix, reply in self.replies.items():
+            if " ".join(command).startswith(prefix):
+                return reply
+        return []
+
+
+class TestBtrfs(unittest.TestCase):
+    def test_it_creates_a_read_only_snapshot_inside_the_subject(self) -> None:
+        run = Runner()
+        with volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=run):
+            pass
+        self.assertEqual(
+            run.calls[0],
+            [
+                "btrfs",
+                "subvolume",
+                "snapshot",
+                "-r",
+                "/var/lib/docker",
+                "/var/lib/docker/.baudolo-20260731",
+            ],
+        )
+
+    def test_it_removes_the_snapshot_afterwards(self) -> None:
+        run = Runner()
+        with volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=run):
+            pass
+        self.assertEqual(
+            run.calls[-1],
+            ["btrfs", "subvolume", "delete", "/var/lib/docker/.baudolo-20260731"],
+        )
+
+    def test_it_maps_a_volume_path_into_the_snapshot(self) -> None:
+        run = Runner()
+        with volume_snapshot(
+            "btrfs", "/var/lib/docker", "20260731", run=run
+        ) as resolve:
+            self.assertEqual(
+                resolve("/var/lib/docker/volumes/postgres_data/_data"),
+                "/var/lib/docker/.baudolo-20260731/volumes/postgres_data/_data",
+            )
+
+    def test_it_keeps_the_trailing_slash_rsync_reads_as_contents(self) -> None:
+        run = Runner()
+        with volume_snapshot(
+            "btrfs", "/var/lib/docker", "20260731", run=run
+        ) as resolve:
+            self.assertEqual(
+                resolve("/var/lib/docker/volumes/postgres_data/_data/"),
+                "/var/lib/docker/.baudolo-20260731/volumes/postgres_data/_data/",
+            )
+
+    def test_it_removes_the_snapshot_even_when_the_body_raises(self) -> None:
+        run = Runner()
+        with (
+            self.assertRaises(ZeroDivisionError),
+            volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=run),
+        ):
+            raise ZeroDivisionError
+        self.assertEqual(run.calls[-1][:3], ["btrfs", "subvolume", "delete"])
+
+
+class TestZfs(unittest.TestCase):
+    def _run(self) -> Runner:
+        return Runner({"zfs list": ["tank/docker"]})
+
+    def test_it_snapshots_the_dataset_mounted_at_the_subject(self) -> None:
+        run = self._run()
+        with volume_snapshot("zfs", "/var/lib/docker", "20260731", run=run):
+            pass
+        self.assertIn(["zfs", "snapshot", "tank/docker@baudolo-20260731"], run.calls)
+
+    def test_it_destroys_the_snapshot_afterwards(self) -> None:
+        run = self._run()
+        with volume_snapshot("zfs", "/var/lib/docker", "20260731", run=run):
+            pass
+        self.assertEqual(
+            run.calls[-1], ["zfs", "destroy", "tank/docker@baudolo-20260731"]
+        )
+
+    def test_it_maps_a_volume_path_through_the_dot_zfs_directory(self) -> None:
+        run = self._run()
+        with volume_snapshot("zfs", "/var/lib/docker", "20260731", run=run) as resolve:
+            self.assertEqual(
+                resolve("/var/lib/docker/volumes/postgres_data/_data"),
+                "/var/lib/docker/.zfs/snapshot/baudolo-20260731/volumes/postgres_data/_data",
+            )
+
+    def test_an_unmounted_dataset_is_an_error(self) -> None:
+        run = Runner({"zfs list": [""]})
+        with (
+            self.assertRaises(SnapshotError),
+            volume_snapshot("zfs", "/var/lib/docker", "20260731", run=run),
+        ):
+            pass
+
+
+class TestRejections(unittest.TestCase):
+    def test_an_unknown_kind_is_rejected(self) -> None:
+        run = Runner()
+        with (
+            self.assertRaises(SnapshotError),
+            volume_snapshot("ext4", "/var/lib/docker", "20260731", run=run),
+        ):
+            pass
+        self.assertEqual(run.calls, [])
+
+    def test_a_path_outside_the_subject_is_rejected(self) -> None:
+        run = Runner()
+        with (
+            volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=run) as resolve,
+            self.assertRaises(SnapshotError),
+        ):
+            resolve("/etc/passwd")
+
+    def test_the_subject_itself_resolves_to_the_snapshot_root(self) -> None:
+        run = Runner()
+        with volume_snapshot(
+            "btrfs", "/var/lib/docker", "20260731", run=run
+        ) as resolve:
+            self.assertEqual(
+                resolve("/var/lib/docker"), "/var/lib/docker/.baudolo-20260731"
+            )
+
+
+class Busy(Runner):
+    def __call__(self, command: list[str]) -> list[str]:
+        if command[:3] == ["btrfs", "subvolume", "delete"]:
+            raise BackupError("target is busy")
+        return super().__call__(command)
+
+
+class TestRemovalFailure(unittest.TestCase):
+    def test_a_failed_removal_does_not_fail_a_completed_run(self) -> None:
+        with volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=Busy()):
+            pass
+
+    def test_a_failed_removal_does_not_mask_the_body(self) -> None:
+        with (
+            self.assertRaises(ZeroDivisionError),
+            volume_snapshot("btrfs", "/var/lib/docker", "20260731", run=Busy()),
+        ):
+            raise ZeroDivisionError
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_snapshot_faithfulness.py

@@ -0,0 +1,127 @@
+"""Which volumes a snapshot of the subject actually contains.
+
+The failure this guards against is silent: a volume with a backing store of
+its own is present inside the snapshot as an empty directory, so rsync
+succeeds, the generation is stamped complete, and the volume is empty in it.
+"""
+
+from __future__ import annotations
+
+import os
+import tempfile
+import unittest
+from pathlib import Path
+from unittest import mock
+
+from baudolo.backup.snapshot import SnapshotError, snapshot_source, unsnapshotted
+from baudolo.backup.volume import Backing
+
+
+class TestUnsnapshotted(unittest.TestCase):
+    def setUp(self) -> None:
+        self.subject = tempfile.mkdtemp()
+        self.mountpoint = str(Path(self.subject) / "volumes" / "app" / "_data")
+        Path(self.mountpoint).mkdir(parents=True)
+
+    def backing(self, **kwargs) -> Backing:
+        return Backing(kwargs.pop("mountpoint", self.mountpoint), **kwargs)
+
+    def test_a_plain_local_volume_is_captured(self) -> None:
+        self.assertIsNone(unsnapshotted(self.backing(), self.subject))
+
+    def test_a_foreign_driver_is_not(self) -> None:
+        reason = unsnapshotted(self.backing(driver="rexray"), self.subject)
+        self.assertIn("rexray", reason)
+
+    def test_declared_driver_options_are_not(self) -> None:
+        reason = unsnapshotted(
+            self.backing(options={"type": "nfs", "device": ":/exports/app"}),
+            self.subject,
+        )
+        self.assertIn("backing store", reason)
+
+    def test_the_declaration_decides_not_the_mount_table(self) -> None:
+        """Docker unmounts an NFS volume when its last container stops."""
+        with mock.patch.object(os.path, "ismount", return_value=False):
+            reason = unsnapshotted(self.backing(options={"type": "nfs"}), self.subject)
+        self.assertIsNotNone(reason)
+
+    def test_a_volume_without_a_mountpoint_is_not(self) -> None:
+        reason = unsnapshotted(Backing(""), self.subject)
+        self.assertIn("no mountpoint", reason)
+
+    def test_an_own_mount_is_not(self) -> None:
+        with mock.patch.object(os.path, "ismount", return_value=True):
+            reason = unsnapshotted(self.backing(), self.subject)
+        self.assertIn("own mount", reason)
+
+    def test_a_filesystem_boundary_is_not(self) -> None:
+        real = os.stat
+
+        def crossing(path, *args, **kwargs):
+            info = real(path, *args, **kwargs)
+            if os.path.realpath(path) == os.path.realpath(self.mountpoint):
+                return os.stat_result(
+                    (info.st_mode, info.st_ino, info.st_dev + 1, *tuple(info)[3:])
+                )
+            return info
+
+        with mock.patch.object(os, "stat", side_effect=crossing):
+            reason = unsnapshotted(self.backing(), self.subject)
+        self.assertIn("filesystem boundary", reason)
+
+    def test_an_unreadable_mountpoint_is_not(self) -> None:
+        reason = unsnapshotted(
+            self.backing(mountpoint=str(Path(self.subject) / "gone")), self.subject
+        )
+        self.assertIn("could not be read", reason)
+
+
+class TestSnapshotSource(unittest.TestCase):
+    def setUp(self) -> None:
+        self.subject = tempfile.mkdtemp()
+        self.mountpoint = str(Path(self.subject) / "volumes" / "app" / "_data")
+        Path(self.mountpoint).mkdir(parents=True)
+        self.snapshot = str(
+            Path(self.subject) / ".baudolo-tag" / "volumes" / "app" / "_data"
+        )
+        Path(self.snapshot).mkdir(parents=True)
+        self.backing = Backing(self.mountpoint)
+
+    def test_a_captured_volume_reads_from_the_snapshot(self) -> None:
+        source, reason = snapshot_source(
+            lambda path: self.snapshot + "/", self.backing, self.subject
+        )
+        self.assertEqual(source, self.snapshot + "/")
+        self.assertEqual(reason, "")
+
+    def test_an_uncaptured_volume_is_refused_before_the_resolver_runs(self) -> None:
+        def resolve(path):
+            raise AssertionError("must not resolve a volume the snapshot misses")
+
+        source, reason = snapshot_source(
+            resolve, Backing(self.mountpoint, options={"type": "nfs"}), self.subject
+        )
+        self.assertIsNone(source)
+        self.assertIn("backing store", reason)
+
+    def test_a_volume_outside_the_subject_degrades_instead_of_raising(self) -> None:
+        def resolve(path):
+            raise SnapshotError(f"{path} lies outside the snapshot subject")
+
+        source, reason = snapshot_source(resolve, self.backing, self.subject)
+        self.assertIsNone(source)
+        self.assertIn("lies outside", reason)
+
+    def test_a_volume_created_after_the_snapshot_degrades(self) -> None:
+        source, reason = snapshot_source(
+            lambda path: str(Path(self.subject) / "absent") + "/",
+            self.backing,
+            self.subject,
+        )
+        self.assertIsNone(source)
+        self.assertIn("created after", reason)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/backup/test_volume.py

@@ -0,0 +1,95 @@
+"""Contract of the rsync invocation that copies a volume."""
+
+from __future__ import annotations
+
+import tempfile
+import unittest
+from pathlib import Path
+from unittest import mock
+
+from baudolo.backup import volume as mod
+
+
+class TestBackupVolume(unittest.TestCase):
+    def copy(self, **kwargs) -> str:
+        with tempfile.TemporaryDirectory() as tmp:
+            defaults = {
+                "versions_dir": tmp,
+                "volume_name": "demo",
+                "volume_dir": str(Path(tmp) / "gen" / "demo"),
+                "authoritative": False,
+                "source": "/var/lib/docker/volumes/demo/_data/",
+            }
+            defaults.update(kwargs)
+            with mock.patch.object(mod, "execute_shell_command") as run:
+                mod.backup_volume(
+                    defaults.pop("versions_dir"),
+                    defaults.pop("volume_name"),
+                    defaults.pop("volume_dir"),
+                    **defaults,
+                )
+            return run.call_args[0][0]
+
+    def test_the_quick_check_pass_carries_no_checksum(self) -> None:
+        self.assertNotIn("--checksum", self.copy(authoritative=False))
+
+    def test_the_authoritative_pass_compares_by_content(self) -> None:
+        self.assertIn("--checksum", self.copy(authoritative=True))
+
+    def test_it_reads_from_the_given_source(self) -> None:
+        command = self.copy(source="/snapshot/volumes/demo/_data/")
+        self.assertIn("/snapshot/volumes/demo/_data/", command)
+
+    def test_it_always_deletes_what_the_source_no_longer_has(self) -> None:
+        self.assertIn("--delete", self.copy())
+
+    def test_it_carries_no_kernel_objects_into_a_generation(self) -> None:
+        self.assertIn("--no-D", self.copy())
+
+    def test_it_keeps_no_twin_of_what_the_second_pass_replaces(self) -> None:
+        command = self.copy(authoritative=True)
+        self.assertEqual(command[:2], ["rsync", "-aP"])
+        self.assertNotIn("--backup", command)
+
+    def test_it_creates_the_destination(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            dest = Path(tmp) / "gen" / "demo"
+            with mock.patch.object(mod, "execute_shell_command"):
+                mod.backup_volume(
+                    tmp, "demo", str(dest), authoritative=False, source="/src/"
+                )
+            self.assertTrue((dest / "files").is_dir())
+
+    def test_source_is_required(self) -> None:
+        with self.assertRaises(TypeError):
+            mod.backup_volume("/v", "demo", "/d", authoritative=False)
+
+    def test_authoritative_is_required(self) -> None:
+        with self.assertRaises(TypeError):
+            mod.backup_volume("/v", "demo", "/d", source="/src/")
+
+
+class TestLastBackupDir(unittest.TestCase):
+    def test_it_ignores_the_generation_being_written(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            current = Path(tmp) / "20260101" / "demo" / "files"
+            current.mkdir(parents=True)
+            found = mod.get_last_backup_dir(tmp, "demo", str(current) + "/")
+            self.assertIsNone(found)
+
+    def test_it_finds_the_previous_generation(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            older = Path(tmp) / "20260101" / "demo" / "files"
+            older.mkdir(parents=True)
+            current = Path(tmp) / "20260102" / "demo" / "files"
+            current.mkdir(parents=True)
+            found = mod.get_last_backup_dir(tmp, "demo", str(current) + "/")
+            self.assertEqual(found, str(older) + "/")
+
+    def test_a_first_run_has_no_predecessor(self) -> None:
+        with tempfile.TemporaryDirectory() as tmp:
+            self.assertIsNone(mod.get_last_backup_dir(tmp, "demo", f"{tmp}/x/"))
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/restore/test_cli_version_flag.py

@@ -0,0 +1,54 @@
+import unittest
+from unittest.mock import patch
+
+from baudolo.restore import __main__ as cli
+
+ENGINES = {
+    "postgres": ("restore_postgres_sql", ["--db-name", "app"]),
+    "mariadb": ("restore_mariadb_sql", ["--db-name", "app"]),
+    "cluster": ("restore_cluster_sql", ["--instance", "central", "--db-user", "root"]),
+}
+
+
+class TestVersionFlagReachesEveryEngine(unittest.TestCase):
+    def call(self, engine: str, extra: list) -> dict:
+        target, required = ENGINES[engine]
+        argv = [
+            engine,
+            "app_vol",
+            "hash",
+            "20260817000000",
+            "--repo-name",
+            "repo",
+            "--container",
+            "db",
+            "--db-password",
+            "pw",
+            *required,
+            *extra,
+        ]
+        with patch.object(cli, target) as restore:
+            self.assertEqual(cli.main(argv), 0)
+        return restore.call_args.kwargs
+
+    def test_the_gate_is_on_by_default(self) -> None:
+        for engine in ENGINES:
+            with self.subTest(engine=engine):
+                self.assertTrue(self.call(engine, [])["check_version"])
+
+    def test_the_flag_turns_it_off(self) -> None:
+        for engine in ENGINES:
+            with self.subTest(engine=engine):
+                kwargs = self.call(engine, ["--no-version-check"])
+                self.assertFalse(kwargs["check_version"])
+
+    def test_empty_stays_independent_of_the_gate(self) -> None:
+        for engine in ENGINES:
+            with self.subTest(engine=engine):
+                kwargs = self.call(engine, ["--empty"])
+                self.assertTrue(kwargs["empty"])
+                self.assertTrue(kwargs["check_version"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/restore/test_cluster_preclean_scope.py

@@ -0,0 +1,126 @@
+"""The pre-clean may only drop what the dump can bring back.
+
+A shared instance carries databases and roles from other applications, and a
+database created after the backup is in no dump at all. Dropping those would
+destroy data this restore cannot restore.
+"""
+
+import tempfile
+import unittest
+from pathlib import Path
+
+from baudolo.restore.db import cluster as cluster_mod
+
+DUMP = """--
+-- PostgreSQL database cluster dump
+--
+
+CREATE ROLE app;
+ALTER ROLE app WITH LOGIN;
+CREATE ROLE reporting;
+
+CREATE DATABASE appdb OWNER app;
+
+\\connect appdb
+
+CREATE TABLE t (v text);
+
+\\connect template1
+"""
+
+
+def dump_file(text: str) -> str:
+    path = Path(tempfile.mkdtemp()) / "central.cluster.backup.sql"
+    path.write_text(text, encoding="utf-8")
+    return str(path)
+
+
+class TestDumpInventory(unittest.TestCase):
+    def test_it_reads_databases_and_roles_the_dump_recreates(self) -> None:
+        databases, roles = cluster_mod.dump_inventory(dump_file(DUMP))
+        self.assertEqual(databases, ["appdb", "template1"])
+        self.assertEqual(roles, ["app", "reporting"])
+
+    def test_a_quoted_name_keeps_its_spaces(self) -> None:
+        databases, roles = cluster_mod.dump_inventory(
+            dump_file('\\connect "odd name"\nCREATE ROLE "odd role";\n')
+        )
+        self.assertEqual(databases, ["odd name"])
+        self.assertEqual(roles, ["odd role"])
+
+    def test_psql_options_are_not_mistaken_for_the_database(self) -> None:
+        databases, _roles = cluster_mod.dump_inventory(
+            dump_file("\\connect -reuse-previous=on dbname=appdb\n")
+        )
+        self.assertEqual(databases, ["appdb"])
+
+    def test_create_database_options_are_not_part_of_the_name(self) -> None:
+        databases, _roles = cluster_mod.dump_inventory(
+            dump_file("CREATE DATABASE appdb WITH TEMPLATE = template0 OWNER = app;\n")
+        )
+        self.assertEqual(databases, ["appdb"])
+
+    def test_a_name_is_listed_once(self) -> None:
+        databases, _roles = cluster_mod.dump_inventory(
+            dump_file("\\connect a\n\\connect a\n")
+        )
+        self.assertEqual(databases, ["a"])
+
+
+class TestInstanceRefusal(unittest.TestCase):
+    """--empty wipes the whole instance, so it may only run on one this dump
+    can rebuild. Scoping the sweep instead wedges the restore: a surviving
+    database that grants to a dumped role pins it, DROP ROLE fails, and the
+    pre-clean aborts after the dump's own databases are already gone."""
+
+    def check(self, present: str, dump: str = DUMP):
+        from unittest import mock
+
+        with mock.patch.object(
+            cluster_mod, "instance_databases", return_value=present.split()
+        ):
+            cluster_mod.assert_instance_matches_dump(
+                "db", "postgres", dump_file(dump), {}
+            )
+
+    def test_an_instance_the_dump_covers_passes(self) -> None:
+        self.check("appdb")
+
+    def test_an_empty_instance_passes(self) -> None:
+        self.check("")
+
+    def test_a_database_the_dump_lacks_is_refused(self) -> None:
+        with self.assertRaises(RuntimeError) as raised:
+            self.check("appdb sibling")
+        self.assertIn("sibling", str(raised.exception))
+
+    def test_the_refusal_names_every_foreign_database(self) -> None:
+        with self.assertRaises(RuntimeError) as raised:
+            self.check("one two")
+        message = str(raised.exception)
+        self.assertIn("one", message)
+        self.assertIn("two", message)
+
+    def test_the_refusal_happens_before_anything_is_dropped(self) -> None:
+        from unittest import mock
+
+        with (
+            mock.patch.object(
+                cluster_mod, "instance_databases", return_value=["foreign"]
+            ),
+            mock.patch.object(cluster_mod, "docker_exec") as touched,
+            self.assertRaises(RuntimeError),
+        ):
+            cluster_mod.restore_cluster_sql(
+                container="db",
+                user="postgres",
+                password="pw",
+                sql_path=dump_file(DUMP),
+                empty=True,
+                check_version=False,
+            )
+        touched.assert_not_called()
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/unit/restore/test_cluster_replay.py

@@ -0,0 +1,125 @@
+import tempfile
+import unittest
+from unittest.mock import MagicMock, patch
+
+from baudolo.restore.db import cluster as cluster_mod
+from baudolo.restore.paths import BackupPaths
+
+
+class TestClusterReplay(unittest.TestCase):
+    def _replay(self, *, empty: bool):
+        calls = []
+
+        def _capture(container, argv, **kwargs):
+            if "-tAc" in argv:
+                return MagicMock(stdout=b"")
+            calls.append((argv, kwargs.get("stdin")))
+            return MagicMock()
+
+        with tempfile.NamedTemporaryFile(suffix=".sql") as sql:
+            sql.write(b"CREATE ROLE app;\nCREATE DATABASE app OWNER app;\n")
+            sql.flush()
+            with patch.object(cluster_mod, "docker_exec", side_effect=_capture):
+                cluster_mod.restore_cluster_sql(
+                    container="db",
+                    user="postgres",
+                    password="pw",
+                    sql_path=sql.name,
+                    empty=empty,
+                    check_version=False,
+                )
+        return calls
+
+    def test_the_replay_is_not_wrapped_in_a_transaction(self) -> None:
+        argv, _ = self._replay(empty=False)[0]
+        self.assertNotIn(
+            "--single-transaction",
+            argv,
+            "CREATE DATABASE cannot run inside a transaction block, so unlike the "
+            "single-database replay this stream must not be wrapped in one",
+        )
+        self.assertIn("ON_ERROR_STOP=1", argv)
+
+    def test_the_replay_targets_the_control_database(self) -> None:
+        argv, _ = self._replay(empty=False)[0]
+        self.assertEqual(argv[argv.index("-d") + 1], cluster_mod.CONTROL_DB)
+        self.assertEqual(argv[argv.index("-U") + 1], "postgres")
+
+    def test_without_empty_nothing_is_dropped_first(self) -> None:
+        self.assertEqual(len(self._replay(empty=False)), 1)
+
+    def test_empty_drops_databases_before_their_owners(self) -> None:
+        calls = self._replay(empty=True)
+        self.assertEqual(len(calls), 2, f"expected pre-clean + replay: {calls}")
+        preclean = calls[0][1].decode()
+        self.assertLess(
+            preclean.index("DROP DATABASE"),
+            preclean.index("DROP ROLE"),
+            "a role cannot be dropped while it still owns a database",
+        )
+        self.assertIn("DROP OWNED BY", preclean)
+        self.assertIn("ORDER BY phase", preclean)
+
+    def test_the_preclean_spares_what_no_dump_recreates(self) -> None:
+        preclean = self._replay(empty=True)[0][1].decode()
+        self.assertIn("NOT datistemplate", preclean)
+        self.assertIn("datname <> current_database()", preclean)
+        self.assertIn("starts_with(rolname, 'pg_')", preclean)
+        self.assertIn("rolname <> current_user", preclean)
+
+    def test_only_the_connecting_role_loses_its_create(self) -> None:
+        # Captured from pg_dumpall 17.
+        dump = [
+            b"CREATE ROLE app;\n",
+            b"ALTER ROLE app WITH NOSUPERUSER INHERIT LOGIN PASSWORD 'SCRAM-SHA-256$...';\n",
+            b"CREATE ROLE postgres;\n",
+            b"ALTER ROLE postgres WITH SUPERUSER INHERIT LOGIN PASSWORD 'SCRAM-SHA-256$...';\n",
+            b'CREATE ROLE "odd-name";\n',
+        ]
+        kept = list(cluster_mod.filter_own_role_creation(dump, "postgres"))
+        self.assertNotIn(b"CREATE ROLE postgres;\n", kept)
+        self.assertIn(b"CREATE ROLE app;\n", kept)
+        self.assertIn(b'CREATE ROLE "odd-name";\n', kept)
+        self.assertEqual(
+            sum(1 for line in kept if line.startswith(b"ALTER ROLE postgres")),
+            1,
+            "the ALTER re-applies the superuser's attributes and password",
+        )
+
+    def test_a_quoted_connecting_role_is_matched_too(self) -> None:
+        kept = list(
+            cluster_mod.filter_own_role_creation(
+                [b'CREATE ROLE "odd-name";\n'], "odd-name"
+            )
+        )
+        self.assertEqual(kept, [])
+
+    def test_a_role_whose_name_merely_starts_the_same_is_kept(self) -> None:
+        kept = list(
+            cluster_mod.filter_own_role_creation(
+                [b"CREATE ROLE postgresql;\n"], "postgres"
+            )
+        )
+        self.assertEqual(kept, [b"CREATE ROLE postgresql;\n"])
+
+    def test_a_missing_dump_is_reported_as_such(self) -> None:
+        with self.assertRaises(FileNotFoundError):
+            cluster_mod.restore_cluster_sql(
+                container="db",
+                user="postgres",
+                password="pw",
+                sql_path="/nonexistent/x.cluster.backup.sql",
+                empty=False,
+                check_version=False,
+            )
+
+    def test_the_path_helper_names_the_dumpall_file(self) -> None:
+        paths = BackupPaths("vol", "hash", "v1", repo_name="repo", backups_dir="/B")
+        self.assertEqual(
+            paths.cluster_file("bigbluebutton"),
+            "/B/hash/repo/v1/vol/sql/bigbluebutton.cluster.backup.sql",
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()