Importing an Immich DB backup into a new CNPG database

This is the alternate Immich DB restore path — for when you have an Immich-format SQL dump (from Immich's built-in backup) rather than a Barman-cloud backup. For the standard Barman recovery path used after a cluster rebuild, see Cluster Rebuild; for the offsite-S3 path, see Offsite Recovery.

Procedure

1. Create the target database:

kubectl cnpg -n databases psql <cluster-name> -- -c 'CREATE DATABASE immich;'

1. Confirm you're against the primary (writes go to the primary; replicas are read-only):

kubectl cnpg -n databases status <cluster-name> | grep 'Primary instance:'

1. Stream the dump in via the Immich restore command. The sed rewrite is required because Immich uses extensions whose search_path needs to be set explicitly during restore:

gunzip < "immich-db-backup-1735455600013.sql.gz" \
  | sed "s/SELECT pg_catalog.set_config('search_path', '', false);/SELECT pg_catalog.set_config('search_path', 'public, pg_catalog', true);/g" \
  | kubectl cnpg -n databases psql <cluster-name> --

1. Verify the row counts after restore (replace <dump-row-count> with what you'd expect from the source):

kubectl cnpg -n databases psql <cluster-name> -- -d immich -c 'SELECT COUNT(*) FROM assets;'

When to use this vs the Barman path

Situation	Use
Cluster rebuild, Barman backups in Garage	Cluster Rebuild → CNPG recovery from Garage
Total loss, only S3 offsite backup survives	Offsite Recovery → Step 4: Restore the database via Barman
Have a pg_dump from inside Immich (admin → "Job Status" → backup)	This doc.
Migrating Immich data between environments	This doc.