Jobs re-run from scratch when trying to scale to 2 machines
Hi,
I have a machine that is already running immich. I use it exclusively for external libraries (i.e. I don't upload anything using immich).
All jobs have completed successfully for this instance (i.e. thumbnails generated, smart search complete, ....).
If I trigger a rescan on external libraries, no new jobs are queued.
Now, I am trying to spawn immich on another machine to scale up.
However, when I do this and trigger a rescan on external libraries, immich queues almost all images and videos for all jobs again.
If I pause/cancel these new jobs, stop the instance on the new machine, and trigger a re-scan on the older machine again, it queues up some jobs (mostly looks like the ones that were completed by the new machine before I stopped them).
The newer machine mounts the
The
I am not sure why all thumbnails need to be generated again (replacing the older thumbnails). Am I doing something dumb?
Will add my
EDIT: Solved. Problem was SSHFS not supporting nanoseconds in file modification time stamps, causing a mismatch, which resulted in immich thinking the file had changed.
I have a machine that is already running immich. I use it exclusively for external libraries (i.e. I don't upload anything using immich).
All jobs have completed successfully for this instance (i.e. thumbnails generated, smart search complete, ....).
If I trigger a rescan on external libraries, no new jobs are queued.
Now, I am trying to spawn immich on another machine to scale up.
However, when I do this and trigger a rescan on external libraries, immich queues almost all images and videos for all jobs again.
If I pause/cancel these new jobs, stop the instance on the new machine, and trigger a re-scan on the older machine again, it queues up some jobs (mostly looks like the ones that were completed by the new machine before I stopped them).
The newer machine mounts the
./library (default UPLOAD LOCATION), and my external libraries via sshfs -o allow_other old_machine:FOLDER NEW-MACHINE-FOLDER, which is then mounted inside docker using typical volume mount.The
docker-compose.ymlredis and database services. 2) immich_server does not have a depends_on (since no redis/database service).I am not sure why all thumbnails need to be generated again (replacing the older thumbnails). Am I doing something dumb?
Will add my
docker-compose.yml.env in a sub-message (character limit issues in this post)EDIT: Solved. Problem was SSHFS not supporting nanoseconds in file modification time stamps, causing a mismatch, which resulted in immich thinking the file had changed.
