immich-server using lots of RAM while idle

[cleoold]

The bug

I did a scan for my external library (around 10K files including RAW) and waited for the jobs to complete. Thumbnails, faces, and places looked fine. The only thing I noticed was it was using 7GB of ram while sitting idle since then. I restarted the container, and mem usage was not that high. Wonder if this could be a memory leak.

After restart:

FYI, this is cockpit dashboard. htop tells the same data.

The OS that Immich Server is running on

Ubuntu 24.04

Version of Immich Server

v1.125.7

Version of Immich Mobile App

NA

Platform with the issue

• Server
• Web
• Mobile

Your docker-compose.yml content

Identical to https://github.com/immich-app/immich/blob/v1.125.7/docker/docker-compose.yml with external library and userns_mode: "keep-id"

Your .env content

Identical to https://github.com/immich-app/immich/blob/v1.125.7/docker/example.env

Reproduction steps

1. run podman-compose up -d
2. create account, add external library and scan
3. Wait for job to complete
4. Observe memory usage not going down
   ...

Relevant log output

Additional information

The only thing different from the default config was I ticked "Prefer embedded preview".

[alextran1502]

Hmm, I checked with docker stats and it doesn't seem to inflate. I wonder if there is something different with docker vs podman here 🤔

[bo0tzz]

It's relatively normal for a process to hold onto some RAM it no longer needs, but once the system asks for it back it should release it. Does it still stay at that usage even if you run something else that needs that RAM?

[cleoold]

Interesting point I've not thought about that. I get some more pictures on my end, let immich scan more and watch what happens.

…

On Sun., Feb. 2, 2025, 5:11 a.m. bo0tzz, ***@***.***> wrote: It's relatively normal for a process to hold onto some RAM it no longer needs, but once the system asks for it back it should release it. Does it still stay at that usage even if you run something else that needs that RAM? — Reply to this email directly, view it on GitHub <#15845 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADKGVB3P5BZUGFW2NV4JAKT2NXVNLAVCNFSM6AAAAABWKA774GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDMMRZGMZDSMZVGA> . You are receiving this because you authored the thread.Message ID: ***@***.***>