mirror of
https://github.com/jellyfin/jellyfin.git
synced 2026-07-25 06:03:37 +03:00
Excessive ffprobe memory usage in jellyfin 10.11.x #8043
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @yeahthegoys on GitHub (Jan 18, 2026).
Description of the bug
When 10.11 had its first stable release public, I, like many people, upgraded to it. I found almost right away that when running a library scan the memory usage on the system would steadily climb until all the memory was used, and at that point the entire VM would crash. At this point I had 8GB of vram assigned to the VM. I tried increasing the allocation to 16GB, the same thing occurs. I tried 32GB. Same thing.
Observing the memory usage I could see it was ffprobe solely responsible for this problem. Further, CPU is almost entirely consumed by ffprobe while this is happening.
I tried nuking my config and running a fresh instance of Jellyfin (at the time, 10.11.2) and letting it build everything from scratch, keeping only the media library (just video files), to rule out a DB migration issue or something of the like. This made no difference. I tried disabling all non-default add-ins. This made no difference.
I revert jellyfin to 10.10.7 using a copy of my old config. Issue is immediately resolved.
I figure no worries, still early days, they'll probably fix it. I forget about it.
Cut to a couple of days ago. I start jellyfin on my TV and the app displays a prominent warning that if I dont upgrade my instance the app will stop working on the next release. Great.
I try the latest release which is now 10.11.5, and lo and behold the same problem is happening.
I have resorted to putting a fixed memory limit of 10GB on the jellyfin container in docker to stop it from taking down the entire vm every 20 minutes or so.
I'm out of options. There are no errors in the log, its not an upgrade issue as I've tried a fresh config multiple times. What was a stable configuration on 10.10.7 is now broken in 10.11.x
Hoping someone can shed some light on this. Thanks in advance.
The included log output begins at a fresh startup of the server and by the end of the log, the issue has started and the memory has already climbed to around 8GB used, and is still climbing.
Reproduction steps
What is the current bug behavior?
Memory usage of ffprobe climbs until all available memory is used. CPU usage by the same process is also maxed out during this time.
What is the expected correct behavior?
Well... not using all available memory and CPU and crashing the host.
Jellyfin Server version
10.11.5
Specify commit id
No response
Specify unstable release number
No response
Specify version number
No response
Specify the build version
10.11.5
Environment
Jellyfin logs
FFmpeg logs
Client / Browser logs
Will provide on request if relevant. Given this happens regardless of whether I have the interface open or not I doubt it.
Relevant screenshots or videos
Red graph is jellyfin container. Usage drops when docker kills ffprobe.
Additional information
No response
@gmipf commented on GitHub (Jan 19, 2026):
This is the exact reason why i had to downgrade back to 10.10.7. i didn't knew what caused it though. Jellyfin is running bare metal, system mem 32GiB. After the crashes I even setup a swap file of 128GiB. Still the same, while scanning libraries it will consume all memory very fast until the system completely crashes.
@nx6 commented on GitHub (Jan 20, 2026):
Do you have music libraries on your server by any chance? I've been working on transitioning apps to a VM from a different setup, and kept having issues with randomly finding the virtual machine not working (crashed I assume). I was going from Jellyfin 10.10.7 on the old setup to making a new 10.11 server (not migrating anything) under Docker on the new VM. I've recently tried again for the third time now, and things are working better. One difference is I have only done the video libraries, abstaining from adding my music to the new server.
@keith-v-smith commented on GitHub (Jan 21, 2026):
Try disabled LUFS.
@yeahthegoys commented on GitHub (Jan 23, 2026):
The only media I have is movies and TV shows. All re-encoded into hevc by tdarr.
@DrHurt commented on GitHub (Jan 24, 2026):
I can confirm ffprobe is a nightmare on my system that goes stray using 400% cpu and even quitting the server doesn't stop it. (On macOS).
I have to force quit ffprobe and restart Jellyfin.
@gnattu commented on GitHub (Jan 24, 2026):
That usually means some file triggered an ffprobe bug. If any of you encountering this could bisect the library and find the file it would be more helpful. We can do little if we don't know what kind of file would trigger this issue.
@yeahthegoys commented on GitHub (Jan 25, 2026):
Noticed this today. See in the graph usage starts climbing at 11.17am. At the same timestamp in the logs I see the below. Previous to this the last time anything was written to log was 11.11am so I'm confident this is related:
Not sure if trickplay uses ffprobe at all? Hope this helps.