Closed tnys closed 11 months ago
Oh wow. The container should normally save everything under the model cache volume at /cache
, so it's strange to see this.
If you haven't deleted the container, would you be able to inspect where all this disk usage is coming from? /tmp
, ~/
and /usr/src/app
are the first places I'd check.
I checked disk usage in the container and these are the results:
root@fd51e8b41d73:/# du -sk /usr
132168 /usr
root@fd51e8b41d73:/# du -sk /cache
2255776 /cache
root@fd51e8b41d73:/# du -sk /tmp
51619696 /tmp
root@fd51e8b41d73:/# du -sk ~/
60 /root/
root@fd51e8b41d73:/# du -sk /usr/src/app
84 /usr/src/app
so it seems the tmp folder seems to be the one causing the issue.
Hmm, my first thought was that model files are downloaded to /tmp, and interrupted downloads led to more files being downloaded and not getting cleaned up. But it would take quite a few retries and large models for that to add up to over 50gb, implausible given your model cache volume is only 1gb.
Based on the library size, my guess is that requests are being spooled to /tmp and not getting cleaned up.
Two questions: How big are the JPEG files in your thumbs folder? Spooling should normally only happen if the images are 16MiB or larger, which could maybe happen if you set JPEG resolution to 4K and quality to 100. Comparatively, how big are the files in /tmp?
The bug
After a few hours of scanning an external library with 100GB of pictures (JPEG, HEIC, MOV, ..) immich_machine_learning uses an excessive amount of diskspace. Output of
docker system df -v
shows 54GB!
The OS that Immich Server is running on
Ubuntu 22.04
Version of Immich Server
v1.86.0
Version of Immich Mobile App
v1.86.0
Platform with the issue
Your docker-compose.yml content
Your .env content
Reproduction steps
Additional information
No response