Closed madadam closed 1 year ago
This was actually a bug: #2618
In the next release the "missing" job should work as expected and you can re-process the skipped files as expected.
You are also describing an "enhancement", which is don't process ML-related jobs if the ML container is offline. Alternatively you could also manually pause/resume the corresponding jobs when ML is started/stopped.
@madadam as this issue has been solved by #2618 , could you please close this issue?
I don't have an easy way to verify the fix right now. Will close anyway and in case I encounter more issues, I'll reopen. Thanks!
No problem. I hope it's been solved tho!
The bug
I have the machine learning container running on a different machine than the rest of the stack, but that machine is not always online. I noticed that the facial recognition job was completed (active: 0, waiting: 0) even though the machine learning machine was offline. Clicking the "MISSING" button on the job did nothing. Clicking "ALL" restarts the job but that loses all the previously recognized faces (from the time the ML container was up)
What I think is happening is the facial recognition job runs but when the request to the ML container fails it still ticks that image as processed and moves on to another one. So in the end all the images are treated as processed by the facial recognition even though they are not. So the job thinks it has nothing left to do and considers itself complete.
What I think should happen is that if the ML container is not accessible the facial recognition job should stop and resume only when the machine learning container is online again.
NOTE: This same issue might affect the other ML jobs as well but I haven't checked.
The OS that Immich Server is running on
OSMC (Debian) on Raspberry PI 4 + Ubuntu 22.04 (only the machine learning container)
Version of Immich Server
v1.59.1
Version of Immich Mobile App
not a mobile issue
Platform with the issue
Your docker-compose.yml content
Your .env content
Reproduction steps
Additional information
No response