Files
collective-lists/.claude/skills/deploy/SKILL.md
Oier Bravo Urtasun ae4fd45b99 deploy(prod): pre-deploy DB backup + paired code/DB rollback skill
Every `just deploy` now takes a pg_dumpall to
/opt/colectivo/backups/predeploy-<ts>-<sha>.sql.gz BEFORE rebuilding the
app or applying migrations, aborts the deploy if the dump is < 1 KiB,
and (on success) appends `<iso-ts> \t <sha> \t <backup-file>` to
/opt/colectivo/.deploys.log on ambrosio. Keeps the newest 10 backups
(prunes older predeploy-* files).

New `infra/scripts/rollback-erosi.sh` reads .deploys.log and pairs a
code rollback with a DB restore atomically:

  just rollback-list          # show recent deploys
  just rollback               # roll back to N-1
  just rollback-to <sha>      # roll back to a specific deploy
  just rollback-code          # roll back code only, keep current DB

Rollback safety:
- 5-second abort window.
- Verifies the target SHA exists locally + the backup is still on prod
  (warns if it's been pruned past the 10-deploy window).
- Uses a temporary git worktree so the user's working tree isn't
  disturbed.
- Stops app/auth/rest/realtime/storage before the gunzip|psql restore.
- Rebuilds the app image at the rolled-back SHA with the same GIT_SHA
  build-arg path the deploy uses (so __APP_COMMIT__ in the bundle
  matches the running code).
- Does NOT write a new .deploys.log entry — rollback is intentionally
  not a deploy event; the next `just deploy` is from current HEAD.
- Storage volume (/var/lib/storage user uploads) is NOT rolled back.

Justfile `deploy` recipe repointed from the stale `deploy.sh` (which
referenced GHCR pull) to `deploy-erosi.sh` (the active prod path).

New project-scoped skill: `.claude/skills/deploy/SKILL.md` documents the
flow + safety boundaries for Claude-assisted invocations. `.gitignore`
keeps `.claude/settings.local.json` ignored but tracks `.claude/skills/`.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-18 14:57:25 +02:00

77 lines
4.3 KiB
Markdown

---
name: deploy
description: Ship the colectivo stack to ambrosio (https://erosi.limonia.net) with pre-deploy DB backup, or roll back code + DB to a previous deploy. Trigger when the user asks to deploy, ship, push to prod, roll back, or undo a deploy.
---
# Deploy + rollback skill
Deploys and rollbacks for the production stack on ambrosio. **Every deploy takes a paired DB backup first**, every successful deploy is logged with `(timestamp, git-sha, backup-filename)`, and rollback uses that log to restore both code and DB atomically.
## Trigger
Use this skill when the user says any of:
- "deploy", "deploy to ambrosio", "deploy to prod", "ship it"
- "roll back", "rollback", "undo the deploy", "revert prod"
- "list deploys", "what was the last deploy"
## What the skill does
### Forward (deploy)
1. **`just deploy`** runs `infra/scripts/deploy-erosi.sh`:
- Captures the local short SHA (`git rev-parse --short HEAD`).
- rsyncs the working tree to `ambrosio:/opt/colectivo/` (excludes `.git`, `node_modules`, `**/.env`, etc.).
- On a fresh `.env`, generates secrets and bails — operator fills `PUBLIC_KEYCLOAK_URL` + `KEYCLOAK_CLIENT_SECRET` and re-runs.
- **Pre-deploy DB backup**: `pg_dumpall` from the prod `db` container → `backups/predeploy-<ts>-<sha>.sql.gz` on ambrosio. Aborts the deploy if the dump is < 1 KiB. Keeps the newest 10 backups.
- Rebuilds the `app` image with `GIT_SHA` build-arg (so `__APP_COMMIT__` lands in the bundle, Fase 14.3).
- `docker compose up -d` brings the stack to the new image.
- Applies any pending DB migrations.
- On success, appends `<iso-ts>\t<sha>\t<backup-file>` to `/opt/colectivo/.deploys.log`.
2. Site smoke: `curl -I https://erosi.limonia.net/` should return 200.
### Reverse (rollback)
1. **`just rollback-list`** prints recent rows from `.deploys.log` so the user can pick a target.
2. **`just rollback`** rolls back code + DB to the previous deploy (last but one line of `.deploys.log`).
3. **`just rollback-to <sha>`** rolls back to a specific deploy.
4. **`just rollback-code`** rolls back code only, leaves the DB at current state — useful when the regression is UI-only and the DB schema is compatible.
`infra/scripts/rollback-erosi.sh` enforces:
- Confirms with a 5-second abort window before doing anything destructive.
- Verifies the target SHA is reachable locally (`git cat-file -e`) and that the backup file still exists on ambrosio.
- Materialises the target commit into a temporary `git worktree`, rsyncs from there, never disturbs the user's working tree.
- Stops `app / auth / rest / realtime / storage` before the DB restore so they don't observe inconsistent state.
- `gunzip | psql -v ON_ERROR_STOP=1` restores the dump.
- Rebuilds the app image at the rolled-back SHA and brings the stack up.
- **Does not** write a new `.deploys.log` line — rollback is intentionally not a deploy event.
## Safety boundaries
- The stack lives on ambrosio. The external proxy + TLS terminator are off-stack (`docs/deployment.md`). Neither this skill nor the scripts ever touch host TLS or the external proxy.
- `infra/scripts/deploy-erosi.sh` never runs `sudo`.
- `--no-verify` / `--amend` are never used. Pre-commit hooks must pass.
- Always confirm with the user before invoking `just rollback*` — DB restore is destructive and irreversible.
- Storage volume (`/var/lib/storage`, user uploads) is **not** part of the rollback. Treated as append-only.
## What to tell the user
Before running anything:
- Show the local short SHA + branch + remote.
- For deploy: confirm with the user. Mention "this will take a fresh DB backup before deploying."
- For rollback: show `just rollback-list` output, ask which SHA, confirm destructive nature.
After running:
- Print the new `.deploys.log` tail (or the rollback target).
- Curl-probe the site and report.
## Files this skill touches
- `infra/scripts/deploy-erosi.sh` — deploy, with pre-deploy backup + .deploys.log writing.
- `infra/scripts/rollback-erosi.sh` — rollback, paired code + DB.
- `Justfile``deploy`, `rollback`, `rollback-list`, `rollback-to`, `rollback-code` recipes.
- `docs/deployment.md` — runbook for both flows.
## Anti-trigger
If the user is just asking *what would happen* on deploy (status, dry-run-style questions), DO NOT run anything. Answer from the script content.