Split collection requirements into Galaxy and Git sources and
introduce retry logic with Git fallback to harden CI against
transient Ansible Galaxy API failures.
- Force PYTHON=python3 for test-local-init to avoid missing venv python
- Rename test-local-inventory-init-all target to test-local-init
- Fix JSON quoting for --vars in local deploy scripts
- Add RUNTIME default to group_vars
- Improve debug output for initial host vars
- Update user hint in local run script
https://chatgpt.com/share/697222e1-5edc-800f-b9ab-511d8afc3142
- introduce env.development marker for local dev runtime
- add cli.meta.runtime module with library API and CLI entrypoint
- detect runtime via ACT/GitHub/marker with explicit RUNTIME override
- use env.development as docker-compose env override
- pass RUNTIME=dev into development inventory generation
- add make setup-development helper
https://chatgpt.com/share/69720fd6-a1c8-800f-b1e2-6909aa63643a
- Use docker-up + ci-discover-output for local discover job (same as CI)
- Add strict input resolver (scripts/meta/resolve/input.sh) with enforced env contract
- Ensure ONLY_APP is always defined (may be empty, but never unset)
- Expose resolved inputs as workflow outputs for linear matrix deploy
- Enforce linear matrix execution via max-parallel: 1
- Align act helper scripts with linear matrix semantics
https://chatgpt.com/share/6970ba64-8c2c-800f-bb90-7edbe9168582
The CI distros runner now only passes arguments supported by
cli.deploy.development run. This avoids argparse errors caused by
forwarding build-only flags that the orchestrator does not accept.
https://chatgpt.com/share/697030c4-f60c-800f-94b6-9128f87da57e
- Build matrix now includes variants: full + slim
- Tag slim images with -slim suffix and use Dockerfile target selection
- Use dedicated gha cache scopes per distro+variant
- Update release checks/retag/promote scripts to handle slim images too
https://chatgpt.com/share/69701a4e-b000-800f-be7e-162dcb93b1d2
- add scripts/inventory/resolve.sh to compute INVENTORY_DIR based on runtime context (act/github/local)
- resolve INVENTORY_DIR dynamically in Makefile via resolve.sh
- enforce INVENTORY_DIR usage across all local deploy/test scripts
- propagate INVENTORY_DIR into container and development CLI commands
- remove hardcoded /etc/inventories paths
https://chatgpt.com/share/69700d69-b270-800f-ad50-d9ff486378b9
Add two local test helpers to run full-stack deployments for all discovered apps:
- make test-local-inventory-init-all: bring up dev stack and create inventory
(/etc/inventories/local-full-<type>/servers.yml) including all apps
- make test-local-run-all: run deploy using the pre-initialized inventory
Also mark various scripts as executable (chmod +x) to ensure consistent
behavior across environments.
https://chatgpt.com/share/6970018e-1244-800f-9345-ca1596061729
- Make TEST_DEPLOY_TYPE strictly required in build-test-matrix
- Enforce required env vars in all act/ci/local deploy scripts
- Add deploy type validation everywhere
- Restore and export PYTHONPATH in Makefile
- Remove unused test directory variables
https://chatgpt.com/share/69700042-3fcc-800f-8bf9-b429d5bb7e82
- Move test-act target logic into scripts/tests/deploy/act/all.sh
- Move test-act-app target logic into scripts/tests/deploy/act/app.sh
- Keep Makefile targets as thin delegators
- Preserve existing flags and behavior
This keeps the Makefile minimal and makes act workflows reusable.
https://chatgpt.com/share/696ff4de-3abc-800f-a206-2f3dd5dec170
- Move test-local-cleanup logic into scripts/tests/deploy/local/cleanup.sh
- Move test-local-rapid logic into scripts/tests/deploy/local/rapid.sh
- Simplify Makefile targets to delegate into scripts
- Keep behavior and flags fully backward compatible
This improves maintainability, reuse in CI, and keeps the Makefile thin.
- execute tests inside running infinito container (no CLI wrapper)
- require INFINITO_DISTRO explicitly and show it in banner
- keep Nix accept-flake-config behavior inside container
- pass TEST_PATTERN/TEST_TYPE as container env and set workdir
- fall back to python3 if PYTHON is unset
https://chatgpt.com/share/696ff1d0-aa84-800f-82cb-05f2b1618d05
- Replace static argparse-based CLI in cli.deploy.development with dynamic
module dispatcher compatible with nested command discovery.
- Ensure cli.deploy.development is executable via python -m and root dispatcher.
- Fix unit test path resolution to work locally and in CI.
- Make docker-* Makefile targets depend on install.
- Simplify test runner script and keep Nix flake config compatibility.
This aligns the CLI architecture with integration test discovery
(cli/**/__main__.py) and removes the need for manual subcommand registration.
https://chatgpt.com/share/696fdbbc-b6d8-800f-a8b5-067aa7e38706
- Introduce cli.deploy.development build subcommand as single build SPOT
- Remove legacy compose.build_infinito implementation
- Make up use build-missing logic via build handler
- Add --when-down flag to up (skip startup if stack already running)
- Simplify restart/run to reuse up handler
- Ensure tests use up --when-down to avoid stack interference
- Make test targets depend on build-missing
https://chatgpt.com/share/696fd3c9-99f0-800f-a0ab-33a91058419a
Move repo root detection into scripts/ci/discover.sh so it can be executed
from any working directory. Simplify ci-discover-output to just call the
script and rely on its internal path resolution.
https://chatgpt.com/share/696e3d20-276c-800f-87a7-a7052989ffe4
Restore discover.sh execution on the host to avoid running Docker Compose
from inside the container, and execute only the storage validation logic
inside the infinito CI container.
This fixes "--profile" errors caused by compose-in-compose and keeps the
Python dependency stack isolated inside the container.
- Run scripts/ci/discover.sh directly on host
- Execute sufficient_storage filter via docker compose exec
- Keep CI output and ONLY_APP override unchanged
https://chatgpt.com/share/696e3d20-276c-800f-87a7-a7052989ffe4
- Add best-effort helpers for Postgres and MariaDB database drops
- Avoid connecting to the target DB when dropping Postgres (use template1 for 'postgres')
- Ignore DB drop errors and continue with purge (compose down + rm -rf)
- Protect MariaDB system databases from being dropped
- Update script header to reflect best-effort behavior
- Support purging multiple stacks in one run
- Best-effort detect postgres/mariadb from .env and drop DB if found
- Always run compose down and delete /opt/docker/<stack>/volumes + stack dir
- Skip DB drop with warning when no backend is detected
https://chatgpt.com/share/696cfcd5-2b84-800f-a386-b7d1d38d68b7
- Add host-level AAAA DNS sanity test against CoreDNS
- Fail hard on SERVFAIL (breaks Node/getaddrinfo with EAI_AGAIN)
- Allow empty AAAA responses (IPv4-only setup is valid)
- Keep A-record validation unchanged
This prevents silent IPv6 resolver regressions that break CSP checks.
https://chatgpt.com/share/696abc27-19dc-800f-9ead-6598c2f3c302
sys-svc-docker role as a post-restart validation handler.
This change moves the inner-container DNS validation out of the CI test
suite and into the Docker service lifecycle itself, ensuring that DNS is
verified exactly where it matters: inside the Docker-in-Docker daemon.
Key changes:
- Add roles/sys-svc-docker/files/test-dns.sh:
- Runs inside the DinD container
- Waits for dockerd readiness
- Executes a BusyBox test container with explicit --dns
- Verifies DOMAIN and subdomain resolve to the expected IP
- Fails hard if DNS is broken
- Register DNS self-test as a docker restart handler:
- Runs once per deployment
- Only executed when DOCKER_IN_CONTAINER and MODE_DEBUG are enabled
- Uses ansible.builtin.script with role_path for correct resolution
- Remove inner Docker DNS test from scripts/tests/dns.sh:
- CI test suite now only validates host → CoreDNS → outer container chain
- DinD DNS validation is now part of the Docker service health lifecycle
This ensures that:
- DNS is validated after every Docker restart
- DinD networking failures are detected immediately
- CI and production share the same DNS validation logic
- Docker-in-Docker DNS becomes a first-class health invariant
https://chatgpt.com/share/696a3dda-5c6c-800f-8a16-47c0c604e1e3
kevinveenbirkenbach