Merge pull request 'feat(backup): restore subcommand (db + documents), proven on the sandbox' (#34) from claude/dolibarr-backup-restore into main

ops/backup/README.md

@@ -60,14 +60,22 @@ orchestrator and the scheduled CronJob (see "Automation" below).
 (`erp/prod/db/…` 1.2 MB, `erp/prod/docs/…` 12.5 MB). Proven end-to-end live on the
 sandbox (dump + tar + GCS upload + retention prune).
 
-## Restore (manual, for now)
+## Restore
 
 ```sh
-# DB:    aws s3 cp s3://arcodange-backup/erp/<env>/db/<ts>.dump - | pg_restore -h <host> -U <user> -d <db> --clean
+ops/backup/dolibarr-backup.sh list    --env <e>                       # find the <ts> key
-# docs:  aws s3 cp s3://arcodange-backup/erp/<env>/docs/<ts>.tar.gz - | tar -C /var/www/documents -xzf -
+ops/backup/dolibarr-backup.sh restore --db   <ts>.dump    --env <e> --yes
+ops/backup/dolibarr-backup.sh restore --docs <ts>.tar.gz  --env <e> --yes
 ```
-The sandbox iso-prod refresh (`ops/sandbox/sandbox-lifecycle.sh`) is the natural
+
-restore-drill bench. A `restore` subcommand is wired next.
+**DESTRUCTIVE** (requires `--yes` + an explicit `--env`): scales the app to 0, then
+- **`--db`**: `DROP OWNED BY <owner_role> CASCADE` + `pg_restore --no-owner --role`
+  (same reset mechanics as `ops/sandbox/sandbox-lifecycle.sh`), then scales back;
+- **`--docs`**: clears `/var/www/documents` and untars the archive, then scales back.
+
+The key is the bare `<ts>` filename from `list`; `--db`/`--docs` selects the
+`db/` or `docs/` subpath. Proven on the sandbox: a mutated `MAIN_INFO_SOCIETE_NOM`
+was reverted to the backup's value by `restore --db`.
 
 ## Automation — the CronJob (gated on creds)
 

ops/backup/dolibarr-backup.sh

@@ -50,8 +50,8 @@ while [[ $# -gt 0 ]]; do
 done
 
 case "$ENV" in
-  prod)    NS="erp";         DB="erp" ;;
+  prod)    NS="erp";         DB="erp";         OWNER_ROLE="erp_role" ;;
-  sandbox) NS="erp-sandbox"; DB="erp-sandbox" ;;
+  sandbox) NS="erp-sandbox"; DB="erp-sandbox"; OWNER_ROLE="erp_sandbox_role" ;;
   *) die "--env must be prod|sandbox" ;;
 esac
 PVC="$NS"
@@ -159,13 +159,88 @@ EOF
   cleanup_secret; trap - EXIT
 }
 
+# restore a half from a stored key. DESTRUCTIVE: scales the app to 0, replaces the
+# DB (DROP OWNED BY <owner> CASCADE + pg_restore) or the documents (clear + untar),
+# then scales back. Mirrors ops/sandbox/sandbox-lifecycle.sh's reset mechanics.
+run_restore() {
+  trap cleanup_secret EXIT
+  copy_s3_secret
+  local DEPLOY="$NS"   # release/instance name == namespace (erp / erp-sandbox)
+  log "Restore ${KIND} on '${ENV}' from ${KEY}  (scaling ${DEPLOY} to 0)"
+  kubectl scale deploy "$DEPLOY" -n "$NS" --replicas=0 >/dev/null 2>&1 || true
+  kubectl wait --for=delete pod -l app.kubernetes.io/instance="$NS" -n "$NS" --timeout=120s >/dev/null 2>&1 || true
+
+  local SCRIPT VOLS="[]" MOUNTS="[]"
+  if [[ "$KIND" == "db" ]]; then
+    SCRIPT="$(cat <<EOF
+$PREAMBLE
+S3 cp "s3://$BUCKET/$PREFIX/db/$KEY" /tmp/r.dump
+echo "fetched \$(wc -c < /tmp/r.dump) bytes"
+psql -h "$PGHOST" -U "\$PGUSER" -d "$DB" -v ON_ERROR_STOP=1 -c "DROP OWNED BY $OWNER_ROLE CASCADE;"
+pg_restore -h "$PGHOST" -U "\$PGUSER" -d "$DB" --no-owner --role=$OWNER_ROLE /tmp/r.dump \\
+  && echo "RESTORED db" || echo "restored db (ignorable warnings)"
+EOF
+)"
+  else
+    VOLS="
+        - name: docs
+          persistentVolumeClaim: { claimName: ${PVC} }"
+    MOUNTS="
+            - { name: docs, mountPath: /docs }"
+    SCRIPT="$(cat <<EOF
+$PREAMBLE
+S3 cp "s3://$BUCKET/$PREFIX/docs/$KEY" /tmp/r.tgz
+echo "fetched \$(wc -c < /tmp/r.tgz) bytes"
+rm -rf /docs/* 2>/dev/null || true
+tar -C /docs -xzf /tmp/r.tgz && echo "RESTORED docs"
+EOF
+)"
+  fi
+  local B64; B64="$(b64 "$SCRIPT")"
+  kubectl delete job dolibarr-restore -n "$NS" --ignore-not-found >/dev/null 2>&1 || true
+  kubectl apply -f - >/dev/null <<EOF
+apiVersion: batch/v1
+kind: Job
+metadata: { name: dolibarr-restore, namespace: $NS }
+spec:
+  backoffLimit: 0
+  ttlSecondsAfterFinished: 600
+  template:
+    spec:
+      restartPolicy: Never
+      volumes: ${VOLS}
+      containers:
+        - name: restore
+          image: $PG_IMAGE
+          envFrom:
+            - secretRef: { name: $TMP_S3_SECRET }
+          env:
+            - { name: PGUSER,     valueFrom: { secretKeyRef: { name: vso-db-credentials, key: username } } }
+            - { name: PGPASSWORD, valueFrom: { secretKeyRef: { name: vso-db-credentials, key: password } } }
+          volumeMounts: ${MOUNTS}
+          command: ["/bin/sh","-c"]
+          args: ["echo $B64 | base64 -d | sh"]
+EOF
+  if ! kubectl wait --for=condition=complete job/dolibarr-restore -n "$NS" --timeout=300s >/dev/null 2>&1; then
+    kubectl logs -n "$NS" job/dolibarr-restore 2>&1 | sed 's/^/    /'
+    kubectl scale deploy "$DEPLOY" -n "$NS" --replicas=1 >/dev/null 2>&1 || true
+    die "restore Job did not complete"
+  fi
+  kubectl logs -n "$NS" job/dolibarr-restore | sed 's/^/    /'
+  kubectl delete job dolibarr-restore -n "$NS" --ignore-not-found >/dev/null 2>&1 || true
+  log "Scaling ${DEPLOY} back to 1"
+  kubectl scale deploy "$DEPLOY" -n "$NS" --replicas=1 >/dev/null 2>&1 || true
+  cleanup_secret; trap - EXIT
+  log "Restore complete."
+}
+
 case "$CMD" in
   backup) run_backup ;;
   list)   run_list ;;
   restore)
     [[ -n "$KEY" && -n "$KIND" ]] || die "restore needs --db <key> or --docs <key>"
     [[ "$YES" == "1" ]] || die "restore is DESTRUCTIVE on '$ENV' — re-run with --yes"
-    die "restore: wired in the chart Job (next iteration) — key=$KEY kind=$KIND env=$ENV"
+    run_restore
     ;;
   *) echo "usage: $0 {backup|list|restore} [--env prod|sandbox] [--db|--docs <key>] [--yes]" >&2; exit 2 ;;
 esac