Claude Fable 5
18239e2fa9
All checks were successful
CD / Deploy to Test (push) Successful in 12s
CD / Smoke tests against Test (push) Successful in 1m15s
CD / Promote to Int (push) Successful in 12s
CI / Lint, typecheck, test (push) Successful in 6m19s
CI / Build container images (push) Has been skipped
CI / Auth e2e pack (push) Successful in 8m13s
CI / Import/export fidelity gate (push) Successful in 54s
CI / Lint, typecheck, test (pull_request) Successful in 6m20s
CI / Build container images (pull_request) Successful in 1m12s
CI / Auth e2e pack (pull_request) Successful in 8m24s
CI / Import/export fidelity gate (pull_request) Successful in 58s
CD / Build and push images (push) Successful in 22s
Adds docs/operations/update-runbook.md (obtain, verify by digest, back up, apply, verify, roll back) with the migration behaviour stated explicitly: a failed migration rolls back its own transaction but is recorded in _prisma_migrations and blocks every further migrate deploy (P3009) — including a re-deployed old image — until migrate resolve --rolled-back; semantically irreversible migrations have exactly one way back, the pre-update backup set. No rolling updates on a compose stage. Rehearsed in the isolated environment of #220: regular update to a v2 image set, then a deliberate failed-update (P3018 division by zero, schema change proven rolled back) with image-rollback-alone shown insufficient and the documented recovery executed. Protocol: docs/vs-nfd/98-update-rollback-protokoll.md. ADR 0024 decisions 5+6 recorded as executed; operations handbook and restore runbook updated; plan checkbox P1-3 ticked. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01AUtYMxwTCMHG9mVHnwbFg8
94 lines
4.6 KiB
Markdown
94 lines
4.6 KiB
Markdown
# Restore runbook (ADR 0015, issue #87)
|
||
|
||
How to restore a Dorfteich stage from a nightly backup set — manually in an
|
||
incident, automatically as the monthly drill. The same procedure doubles as
|
||
the **Prod relocation procedure**: restore the latest set on the new host.
|
||
|
||
A restore set is one backup id `YYYYMMDD-HHMMSS`: `db-<id>.dump`
|
||
(`pg_dump -Fc`) plus `files-<id>.tar.gz` (uploads + plugins volumes),
|
||
written nightly by the `backup` sidecar onto the `backups` volume, with
|
||
`status.json` describing the last run (deploy/monitoring.md).
|
||
|
||
## Manual restore (incident / relocation)
|
||
|
||
On the stage host, from the stage directory (`/srv/DOCKER/dorfteich-<stage>/`):
|
||
|
||
1. **Pick the set.** `docker compose exec backup ls /backups` — usually the
|
||
id in `status.json` → `lastSuccess.backupId`.
|
||
2. **Run the automated runbook:** `./restore.sh <backup-id>`
|
||
(`deploy/backup/restore.sh`). It stops `web`/`api`/`collab` (the db stays
|
||
up), resets the `public` schema and replays the dump with `pg_restore`
|
||
(#288 — objects created after the backup do not survive) and unpacks
|
||
the volume archive through the backup sidecar image, starts the stack,
|
||
and polls `/readyz`.
|
||
3. **Verify:** `/readyz` fully green, spot-check one page and one uploaded
|
||
file in the browser.
|
||
|
||
Consistency model (ADR 0015): the volume archive is taken minutes after the
|
||
dump — a page referencing a file uploaded in between shows a missing image,
|
||
never corruption.
|
||
|
||
Restore is also the designated way back from an update whose migrations
|
||
succeeded but must be undone — the pre-update backup set is step 3 of
|
||
`docs/operations/update-runbook.md` (#221).
|
||
|
||
**Relocation to a new host:** provision the stage directory (compose +
|
||
`.env`, deploy/stages.md), start only `db` and `backup`
|
||
(`docker compose up -d db backup`), copy the set into the backups volume
|
||
(`docker run --rm -v <src> -v <project>_backups:/backups …`), then steps 2–3.
|
||
|
||
## In-app restore (issue #103)
|
||
|
||
With a Nextcloud target configured, Site Admins can restore without shell
|
||
access: _Admin → System → Backups → Restore_ lists local and remote sets;
|
||
after a type-to-confirm prompt the backup sidecar orchestrates the whole
|
||
restore (maintenance mode → download + verify → terminate connections →
|
||
`pg_restore` + volume extract → api restart). Progress lands in
|
||
`restore-status.json` next to `status.json`; the public
|
||
`GET /api/v1/backup/restore-status` endpoint keeps answering while
|
||
everything else serves 503 `maintenance_mode`. This runbook stays the
|
||
disaster path for when the app itself is gone.
|
||
|
||
## Fetching a set from Nextcloud (total loss)
|
||
|
||
When the host is gone but the off-host copies exist, rebuild from the
|
||
Nextcloud bundle alone:
|
||
|
||
1. Download `dorfteich-backup-<id>.tar.gz` from the configured Nextcloud
|
||
folder (browser or `curl -u <user>:<app-password> -O
|
||
https://cloud.example.com/remote.php/dav/files/<user>/<folder>/dorfteich-backup-<id>.tar.gz`).
|
||
2. Unpack it: `tar -xzf dorfteich-backup-<id>.tar.gz` → `db-<id>.dump`,
|
||
`files-<id>.tar.gz`, and a `manifest.json` describing the set.
|
||
3. Copy the two artifacts into the (fresh) stack's backups volume — see
|
||
_Relocation to a new host_ above — and run `./restore.sh <id>`.
|
||
|
||
## Automated monthly drill (`.gitea/workflows/drill.yml`)
|
||
|
||
Runs on the 1st of each month — and on demand by pushing a `drill-*` tag
|
||
(`git tag drill-$(date +%s) && git push origin --tags`; Gitea 1.22 has no
|
||
workflow-dispatch button yet). It executes `deploy/backup/drill.sh`, which
|
||
|
||
- reads the drilled stage's backups volume **read-only** (stage volumes are
|
||
never touched — everything scratch lives under a unique
|
||
`dorfteich-drill-<timestamp>` prefix and is removed afterwards),
|
||
- restores the latest successful set into a throwaway Postgres + volumes
|
||
using the same backup-image code path as `restore.sh`,
|
||
- boots the api image against the result and checks: readyz database +
|
||
migrations ok, ≥ 1 user and live page, ≥ 1 rendered page in the content
|
||
cache, a public API request answers 200, and one media file's bytes on
|
||
the volume match its database row,
|
||
- reports the outcome as a comment on the pinned **Restore drills** issue
|
||
(#98), then tears the scratch environment down (also on failure).
|
||
|
||
Pre-go-live the drill restores the **Test** stage's set
|
||
(`DRILL_SOURCE_VOLUME: dorfteich-test_backups`); at go-live (#89) point it
|
||
at the Prod backups volume. Manual invocation on the stage host:
|
||
|
||
```sh
|
||
SOURCE_VOLUME=dorfteich-test_backups sh deploy/backup/drill.sh
|
||
```
|
||
|
||
A drill failure means the current backup set is **not restorable** — treat
|
||
it like a failed backup: check the sidecar logs and `status.json`, fix, and
|
||
re-run the drill the same day.
|