ImmichI
Immich•6mo ago
nkdf.

Facial Recognition hangs

Long term immich user here running on v1.135.3 docker installation. I've noticed that my facial recognition job hangs after a while and now I've got quite a backlog (~30k) images. Turning on debug, I see the following message: 
immich_server            | [Nest] 7  - 07/02/2025, 8:42:54 PM   DEBUG [Microservices:PersonService] Face 7457731c-4087-4549-93de-ad7ab07d5995 only matched the face itself, skipping
immich_server            | Query failed : {
immich_server            |   durationMs: 2.321487999986857,
immich_server            |   error: Error: write CONNECTION_DESTROYED immich_postgres:5432
immich_server            |       at Object.execute (/usr/src/app/node_modules/postgres/cjs/src/connection.js:156:35)
immich_server            |       at Query.handler (/usr/src/app/node_modules/postgres/cjs/src/index.js:229:13)
immich_server            |       at Query.handle (/usr/src/app/node_modules/postgres/cjs/src/query.js:140:65)
immich_server            |       at process.processTicksAndRejections (node:internal/process/task_queues:105:5) {
immich_server            |     code: 'CONNECTION_DESTROYED',
immich_server            |     errno: 'CONNECTION_DESTROYED',
immich_server            |     address: [ 'immich_postgres' ],
immich_server            |     port: [ 5432 ]
immich_server            |   },
immich_server            |   sql: 'begin',
immich_server            |   params: []
immich_server            | }

Resource usage on the server is normal (50%cpu, 20GB ram free), and the database is hosted on local ssd (non-os) with 28GB of free space. There are around 200k assets on this server. Restarting the server or container restarts the job at ~55k assets in queue, debug logs show most of them are non-core faces for later processing.

If I hit clear, the "1" active remains persistent, but the job never resumes until the container is restarted.
Was this page helpful?