mirror of
https://github.com/kevinveenbirkenbach/computer-playbook.git
synced 2026-11-11 05:06:36 +00:00
The NFS backup rsynced the whole export and honoured only one exclude, the shared backup root. Volumes declared `backup: false` were copied anyway and then pulled a second time by remote-2-local; in the swarm DR drill of Magento#0 (run 40625679270) docs_sites alone (a full repo clone plus its prepared and work trees) took about 15.5 GiB per copy and filled the runner. The volume backup already skipped these through the backup_volume lookup; only the NFS path ignored the declaration. - utils/storage/nfs.py: volume_dir() names the state directory a volume binds to (pinned name or key), shared with compose_volumes; backup_disabled_dirs() returns every NFS-backed `backup: false` volume dir that no backed-up volume also binds to. - The role-local lookup backup_nfs_excludes feeds them into BKP_NFS_2_LOCAL_EXCLUDES next to the backup root; script.sh takes any number of excludes. Compose installs still pass only the backup root. - recover.py protects every live state directory the snapshot does not hold, so `rsync --delete` keeps the excluded volumes (model weights, the docs tree, caches) and volumes created after the snapshot instead of wiping them. - The swarm DR drill restores BACKUP_NFS_VOLUME, the first NFS volume not declared `backup: false`; PRIMARY_NFS_VOLUME stays for the drain tests. BookWyrm now drills bookwyrm_media; decidim, fusiondirectory, meta, coturn, svc-registry-cache and svc-registry-docker have nothing backed up on NFS and skip the drill. Measured: compose_volumes renders identically before and after for 291 roles across compose/swarm x local/nfs (1940 of 1940); the 15 excluded directories are exactly the backup-disabled device dirs; the systemd unit renders shell-safe on ansible-core 2.20 and 2.21 and stays byte-identical in compose and swarm-local. A new unit test runs the real script.sh over two generations and the real recover.py back into the export, and fails without the restore protection. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
185 lines
6.2 KiB
Bash
Executable File
185 lines
6.2 KiB
Bash
Executable File
#!/usr/bin/env bash
|
|
set -euo pipefail
|
|
|
|
_REPO_ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")/../../../../.." && pwd)"
|
|
# shellcheck source=scripts/meta/env/load.sh
|
|
source "${_REPO_ROOT}/scripts/meta/env/load.sh"
|
|
|
|
if [ -s "${SWARM_DRILL_ENV:-}" ]; then
|
|
set -a
|
|
# shellcheck source=/dev/null
|
|
source "${SWARM_DRILL_ENV}"
|
|
set +a
|
|
fi
|
|
|
|
: "${APP_ID:?APP_ID required}"
|
|
|
|
ENTITY="$(PYTHONPATH="${_REPO_ROOT}" "${PYTHON}" -c "from utils.roles.entity.name import entity_name; print(entity_name('${APP_ID}'))")"
|
|
|
|
STACK_NAME="${ENTITY}"
|
|
CUSTOM_IMAGE_REPO="${ENTITY}_custom"
|
|
|
|
ROLE_DIR="${_REPO_ROOT}/roles/${APP_ID}"
|
|
|
|
PRIMARY_SERVICE_KEY="$(PYTHONPATH="${_REPO_ROOT}" "${PYTHON}" -c "
|
|
import os, yaml
|
|
key = '${ENTITY}'
|
|
svc = os.path.join('${ROLE_DIR}', 'meta', 'services.yml')
|
|
if os.path.exists(svc):
|
|
data = yaml.safe_load(open(svc)) or {}
|
|
entity = data.get('${ENTITY}') if isinstance(data, dict) else None
|
|
if isinstance(entity, dict):
|
|
k = (entity.get('proxy') or {}).get('service_key')
|
|
if isinstance(k, str) and k:
|
|
key = k
|
|
print(key)
|
|
")"
|
|
SERVICE_NAME="${STACK_NAME}_${PRIMARY_SERVICE_KEY}"
|
|
|
|
DB_DEP=none
|
|
if [ -f "${ROLE_DIR}/meta/services.yml" ]; then
|
|
if grep -qE '^mariadb:' "${ROLE_DIR}/meta/services.yml"; then
|
|
DB_DEP=mariadb
|
|
elif grep -qE '^postgres:' "${ROLE_DIR}/meta/services.yml"; then
|
|
DB_DEP=postgres
|
|
fi
|
|
fi
|
|
|
|
DEFAULT_PLACEMENT_MANAGER="$(PYTHONPATH="${_REPO_ROOT}" "${PYTHON}" -c "
|
|
from utils.roles.meta_lookup import get_role_placement
|
|
placement = get_role_placement('${ROLE_DIR}', role_name='${APP_ID}') or ''
|
|
print('true' if placement.strip() == 'manager' else 'false')
|
|
")"
|
|
|
|
PROBE_PORT="$(PYTHONPATH="${_REPO_ROOT}" "${PYTHON}" -c "
|
|
import os, yaml
|
|
port = 80
|
|
svc = os.path.join('${ROLE_DIR}', 'meta', 'services.yml')
|
|
if os.path.exists(svc):
|
|
data = yaml.safe_load(open(svc)) or {}
|
|
entity = data.get('${ENTITY}') if isinstance(data, dict) else None
|
|
if isinstance(entity, dict):
|
|
try:
|
|
port = entity['ports']['internal']['http'] or 80
|
|
except Exception:
|
|
port = 80
|
|
print(port)
|
|
")"
|
|
|
|
NFS_VOLUMES=""
|
|
BACKUP_NFS_VOLUMES=""
|
|
if [ -f "${ROLE_DIR}/meta/volumes.yml" ]; then
|
|
_nfs_volumes_py="
|
|
import sys, yaml
|
|
with open('${ROLE_DIR}/meta/volumes.yml') as f:
|
|
data = yaml.safe_load(f) or {}
|
|
if isinstance(data, list):
|
|
entries = [e for e in data if isinstance(e, dict)]
|
|
elif isinstance(data, dict):
|
|
entries = [v for v in data.values() if isinstance(v, dict)]
|
|
else:
|
|
entries = []
|
|
for v in entries:
|
|
if v.get('type') != 'volume':
|
|
continue
|
|
if v.get('nfs') is False:
|
|
continue
|
|
if sys.argv[1:] == ['backed-up'] and v.get('backup') is False:
|
|
continue
|
|
docker_name = v.get('docker_name') or v.get('name')
|
|
if isinstance(docker_name, str):
|
|
print(docker_name)
|
|
"
|
|
NFS_VOLUMES="$(python3 -c "${_nfs_volumes_py}")"
|
|
BACKUP_NFS_VOLUMES="$(python3 -c "${_nfs_volumes_py}" backed-up)"
|
|
fi
|
|
|
|
PRIMARY_NFS_VOLUME="${NFS_VOLUMES%%$'\n'*}"
|
|
BACKUP_NFS_VOLUME="${BACKUP_NFS_VOLUMES%%$'\n'*}"
|
|
|
|
NFS_CHECK_MOUNTPOINT="/mnt/nfs-check"
|
|
|
|
export APP_ID ENTITY STACK_NAME SERVICE_NAME PRIMARY_SERVICE_KEY CUSTOM_IMAGE_REPO DB_DEP NFS_VOLUMES PRIMARY_NFS_VOLUME BACKUP_NFS_VOLUME DEFAULT_PLACEMENT_MANAGER PROBE_PORT NFS_CHECK_MOUNTPOINT
|
|
|
|
# Exits the calling chaos step (09/10/11) with success when the role is
|
|
# manager-pinned (no worker task exists to drain).
|
|
skip_chaos_if_manager_pinned() {
|
|
if [ "${DEFAULT_PLACEMENT_MANAGER}" = true ]; then
|
|
echo "SKIP: ${ENTITY} declares placement: manager (single-node on the swarm manager) — drain-a-worker reschedule chaos does not apply"
|
|
exit 0
|
|
fi
|
|
}
|
|
|
|
has_swarm_service() {
|
|
local _names=""
|
|
[ -n "${MGR:-}" ] || return 1
|
|
if ! _names="$(docker exec "${MGR}" docker stack services \
|
|
--format '{{.Name}}' "${STACK_NAME}" 2>/dev/null)"; then
|
|
_names=""
|
|
fi
|
|
printf '%s\n' "${_names}" | grep -qxF "${SERVICE_NAME}"
|
|
}
|
|
|
|
# Prints the swarm service carrying APP_ID's database. A role that pins its own
|
|
# engine runs it as <entity>_database inside its own stack, a central provider
|
|
# as <dep>_<dep> in its own (plugins/lookup/database.py).
|
|
resolve_db_service() {
|
|
local _names=""
|
|
[ "${DB_DEP}" != "none" ] || return 0
|
|
[ -n "${MGR:-}" ] || return 0
|
|
if ! _names="$(timeout 30 docker exec "${MGR}" docker service ls --format '{{.Name}}' 2>/dev/null)"; then
|
|
return 0
|
|
fi
|
|
if printf '%s\n' "${_names}" | grep -qxF "${STACK_NAME}_database"; then
|
|
printf '%s\n' "${STACK_NAME}_database"
|
|
elif printf '%s\n' "${_names}" | grep -qxF "${DB_DEP}_${DB_DEP}"; then
|
|
printf '%s\n' "${DB_DEP}_${DB_DEP}"
|
|
fi
|
|
}
|
|
|
|
# Exits the calling gate (03/05/06/07) with success when the role deploys no
|
|
# swarm service to converge or drain.
|
|
skip_if_no_swarm_service() {
|
|
if ! has_swarm_service; then
|
|
echo "SKIP: ${ENTITY} (${APP_ID}) deployed no swarm service — converge/chaos gate does not apply"
|
|
exit 0
|
|
fi
|
|
}
|
|
|
|
# A stack whose sidecar is named after its app (<app>mcp) makes every
|
|
# `--filter name=<app>` a two-hit match, so both helpers below match exactly.
|
|
# Param: $1 node container, $2 swarm service name
|
|
service_container_id() {
|
|
docker exec "$1" docker ps \
|
|
--filter "label=com.docker.swarm.service.name=$2" \
|
|
--format '{{.ID}}' | head -1
|
|
}
|
|
|
|
# Param: $1 manager container, $2 swarm service name
|
|
service_replicas() {
|
|
docker exec "$1" docker service ls \
|
|
--filter "name=$2" --format '{{.Name}} {{.Replicas}}' |
|
|
awk -v svc="$2" '$1 == svc { print $2 }'
|
|
}
|
|
|
|
# Param: $1 node container, $2 app container id, $3 port
|
|
probe_app_reachable() {
|
|
docker exec "$1" docker exec "$2" sh -c \
|
|
"curl -sS http://localhost:$3/ || wget -qO- http://localhost:$3/ || nc -z localhost $3" >/dev/null 2>&1 ||
|
|
docker exec "$1" docker exec "$2" bash -c "exec 3<>/dev/tcp/localhost/$3" >/dev/null 2>&1 ||
|
|
docker exec "$1" sh -c "docker inspect -f '{{.State.Health.Status}}' $2 | grep -qx healthy" >/dev/null 2>&1
|
|
}
|
|
|
|
if [ "${SWARM_NFS_PILOT_VERBOSE:-0}" = "1" ]; then
|
|
echo "==> APP_ID=${APP_ID}"
|
|
echo " ENTITY=${ENTITY}"
|
|
echo " STACK_NAME=${STACK_NAME}"
|
|
echo " SERVICE_NAME=${SERVICE_NAME}"
|
|
echo " CUSTOM_IMAGE_REPO=${CUSTOM_IMAGE_REPO}"
|
|
echo " DB_DEP=${DB_DEP}"
|
|
echo " PRIMARY_NFS_VOLUME=${PRIMARY_NFS_VOLUME}"
|
|
echo " BACKUP_NFS_VOLUME=${BACKUP_NFS_VOLUME}"
|
|
echo " DEFAULT_PLACEMENT_MANAGER=${DEFAULT_PLACEMENT_MANAGER}"
|
|
echo " PROBE_PORT=${PROBE_PORT}"
|
|
fi
|