Latest Blog

Choose repair or rebuild by persistent-data integrity, failure scope, reproducibility, and the ability to validate a clean recovery target.

Size an Immich reserve from measured temporary and derived output, set byte and inode alerts, and expand before jobs consume the recovery margin.

Decide which Immich paths can use a network share, verify mount ordering and identity, then test restart and NAS-outage behavior.

Diagnose Immich with one fixed workload, resource evidence, and one-variable retests before upgrading hardware or changing concurrency.

Tune a small Immich server around the busiest family upload, then cap concurrency and schedule derived-media work without weakening backup safety.

Prevent upgrade loss by versioning deployment files, backing up every persistent component, staging changes, and testing recreation before cutover.

Recover a failed Immich update with a pinned rollback first, a coordinated restore only if needed, and production checks before the next attempt.

Separate missing photos from missing search results, repair the failed layer, then verify Smart Search across two controlled restarts.