---
name: metabase-versioning
id: 0006
status: pending
created: 2026-05-16
updated: 2026-05-16
priority: medium
risk: medium
related_issues: []
apps:
  - auto_metabase
  - dag_engine
  - agents_and_robots
trigger: cron
schedule: "0 2 * * *"
expected_runtime_s: 60
tags: [metabase, sync, gitops, backup]
---

## Goal

Probar flujo INVERSO al tipico: extraer estado de un servicio interno (Metabase) y persistirlo como codigo. Sirve de backup + auditoria + reproducibilidad.

## Pre-requisitos

- Metabase corriendo (local docker o produccion).
- `auto_metabase` configurado con credenciales (`METABASE_URL`, `METABASE_TOKEN`).
- `projects/metabase_registry/` existe con sub-repo git inicializado.

## Flow

1. DAG `metabase-snapshot.yaml` diario 02:00:
   ```yaml
   name: metabase-snapshot
   schedule: "0 2 * * *"
   steps:
     - name: pull_metabase
       function: auto_metabase_pull_bash_pipelines
       args: ["--target", "projects/metabase_registry"]
     - name: git_diff_check
       command: cd projects/metabase_registry && git diff --stat
     - name: commit_if_changes
       command: |
         cd projects/metabase_registry
         git add -A
         if ! git diff --cached --quiet; then
           git commit -m "snapshot: $(date -u +%FT%TZ)"
           git push origin master
           echo "CHANGES_PUSHED"
         else
           echo "NO_CHANGES"
         fi
       depends: [pull_metabase]
     - name: notify
       function: matrix_send_message_<id>
       args: ["#fn-registry-ops", "Metabase snapshot: <stdout from prev step>"]
       depends: [commit_if_changes]
   ```
2. data_factory: node `metabase_snapshot` kind=extractor (source=metabase). Tag `gitops`.
3. Verificar que YAML files generados son legibles y diff-friendly.

## Acceptance

- [ ] DAG corre 3 dias consecutivos sin error.
- [ ] >=1 commit registrado en repo `metabase_registry`.
- [ ] data_factory.runs muestra historico.
- [ ] Matrix recibe 1 mensaje/dia (con "NO_CHANGES" o "CHANGES_PUSHED").
- [ ] Si se rompe un dashboard manualmente en Metabase -> push de YAML viejo lo restaura (test).

## Telemetria esperada

- 1 run/dia en data_factory.
- 7 commits en metabase_registry repo (1 semana baseline).

## Definition of Done

Ver `README.md` seccion DoD + user-facing.

### Generico

- [ ] **Repetibilidad**: cron diario 02:00 corre 7 dias consecutivos sin error.
- [ ] **Observabilidad**: `data_factory.runs` + 1 commit en repo `metabase_registry` por dia (o `NO_CHANGES`).
- [ ] **Error-path**: token Metabase expirado → healthcheck pre-pull falla con mensaje claro, no silencio.
- [ ] **Idempotencia**: NO_CHANGES no genera commit vacio en git.
- [ ] **Secrets**: token Metabase en `pass metabase/api-token`.
- [ ] **Docs**: `## Notas` con onboarding + rollback procedure.
- [ ] **Registry-first**: pull/diff/push reusan funciones registry.
- [ ] **INDEX + status**: `status: done` + INDEX + movido.

### User-facing

- [ ] **User-facing**: usuario navega a `https://gitea.../dataforge/metabase_registry/commits/master` → ve commits diarios con diff YAML de dashboards/cards.
- [ ] **User-facing repeat**: misma URL manana muestra commit nuevo (o `NO_CHANGES` skip); rollback con click derecho en commit → restore.
- [ ] **User-facing onboarding**: parrafo en `## Notas`: "Para auditar cambios Metabase: abrir Gitea repo `dataforge/metabase_registry`. Rollback: revertir commit en Gitea → push trigger DAG manual → Metabase restaurado. Matrix bot diario en `#fn-registry-ops` a las 09:00."
- [ ] **User-facing latencia**: cambio manual en Metabase → commit visible al dia siguiente 02:00.

### Custom

- [ ] Rollback E2E probado: revertir commit → siguiente run aplica YAML viejo → Metabase restaura dashboard.
- [ ] Diff YAML estable: keys ordenadas, no churn aleatorio.
- [ ] Dashboards eliminados → commit `DELETED:`, no tombstone huerfano.
- [ ] Backup adicional a vault (no solo git).

## Notas

- Riesgo: si Metabase token expira, el DAG falla silenciosamente. Anadir healthcheck pre-pull.
- Restore es manual hoy. Futuro: comando `auto_metabase push --from-commit <SHA>`.