Skip to content

Immich.md

Latest commit

 

History

History
48 lines (35 loc) · 1.43 KB

Immich.md

File metadata and controls

48 lines (35 loc) · 1.43 KB
Raw
title Immich
tags
services
database

Immich

Photo management service. Postgres runs as a primary on hermes with a streaming replica on zeus.

Backups

Daily at 03:00 via the immich-backup periodic job. Dumps are written (gzipped) to the Synology NAS at dionysus.internal:/volume1/data/immich-postgres-backup/ and retained for 7 days.

To trigger a manual backup immediately:

nomad job periodic force immich-backup

To restore from a backup (run from a host with psql access):

gunzip -c /path/to/backup.sql.gz | psql -h <postgres-host> -U <user> -v ON_ERROR_STOP=0

Failover: Promoting the Zeus Replica

If hermes becomes unavailable or its disk dies:

1. Find the replica allocation:

nomad job allocs immich

Note the alloc ID for the backend-replica group.

2. Promote the replica to primary:

nomad alloc exec -task postgres <alloc-id> psql -U <POSTGRES_USER> -d postgres -c "SELECT pg_promote()"

3. Update jobs/immich/immich.hcl — swap the constraints:

  • backend group: value = "hermes" → value = "zeus", change port to match zeus
  • backend-replica group: remove or set to value = "hermes" (once restored)

4. Redeploy:

nomad job run jobs/immich/immich.hcl

5. Once hermes is restored, re-seed it as the new replica: set it up as a host_volume node, update the constraint back, and let replica-init run pg_basebackup automatically on next deploy.