possible memory leak?

[mick-pihole]

Describe the bug
memory keep going up till Synology crash, unreachable
had to reboot Synology.

update: I set memory limit at 2GB, the jellyfin is not crashing, however memory usage stays at 2GB.

System (please complete the following information):

• OS: Synology

• Virtualization: docker

• Clients: both Browser and Fire Stick, have not tested others

• Browser: Chrome

• Jellyfin Version: 10.7.6

• Playback: Transcode or direct stream

• Hardware Acceleration: VAAPI

• Installed Plugins: opensubtitle

• Reverse Proxy: none

• Base URL: [e.g. none, yes: /example]

• Networking: [e.g. Host, Bridge/NAT]

• Storage: [e.g. local, NFS, cloud]

To Reproduce

1. Go to '...'
2. Click on '....'
3. Scroll down to '....'
4. See error

Expected behavior

Server Logs

FFmpeg Logs

Browser Console Logs

Screenshots

Additional context

[josolanes]

I see the same behavior in a docker container run within Gentoo and Jellyfin version 10.7.6

Overnight mine used a little over 6GB of memory (with no limit) when started at 3:30p and checked this morning at 8a. I've since limited its container to 4GB

Mine is a fresh install (as of yesterday) with the following libraries and sizes:

• Movies: 228
• Music: 6031
• Books: 31
• Shows: 1873

Memory usage continued to climb after scanning the libraries and no scheduled tasks were listed as recently running when I noticed it in the morning. Jellyfin was also not in active use (nothing listed in activity overnight)

Mine is behind a reverse proxy (using jwilder/nginx-proxy:latest) and I'm using NVENC hardware acceleration. No new plugins or anything in my case

EDIT: The container limited to 4GB reached the container limit, crashed, and automatically restart itself (per container settings) after 8 hours

[MafiaInc]

I'm not sure whether is memory leak or not, but I've noticed Jellyfin is using about 3.2GB of memory right now so I have decided to search and ended up in this issue. Here are some details about my setup:

Server: 4acdaf640013
Version: 10.7.6
OS: Linux
Architecture: X86

It runs in docker container on VM. It was used maximum by three users at a time. Memory usage :

# smem -P jellyfin -k
  PID User     Command                         Swap      USS      PSS      RSS
380643 root     python /usr/bin/smem -P jel        0    10.5M    10.7M    12.3M
2859268 root     /jellyfin/jellyfin              2.1M     3.2G     3.2G     3.2G

I will keep it on eye to see whether is increasing or not. The process is running since Aug26, so it is three days old.
I've searched the website for some info about the hardware requirements, I couldn't find.

Let me know if you would require more info for the process, I could provide.

Thanks for all the efforts!

[cvium]

It is only a memory leak if the memory usage continues to climb without ever stopping. There's a lot of stuff being cached in memory, so high usage isn't necessarily any indication.

[FilipLac]

I have same issue on:
Version: 10.7.7
Operating System: Linux (based on debian)
Architecture: Arm64
Running on RPI 4b

It cached video that I tried to play in memory and left it there even after completing that video to the end. When I tried playing it on different device it cached it too, and now it looks like it has that video in memory twice. And unfortunately only way to get rid of it is restarting whole device or that dock in docker.

[cvium]

That sounds extremely unlikely. Neither ffmpeg nor jellyfin reads entire video files into RAM.

[FilipLac]

That sounds extremely unlikely. Neither ffmpeg nor jellyfin reads entire video files into RAM.

Well all I know that I had there 200MB video and when I played it 2 times my RAM usage on service Jellyfin went up by same amount.

[lordkitsuna]

Just throwing my issue into here. i can get this issue with absolutely no use of jellyfin whatsoever. i just have the server running and over a few days it will end up sigkilled when the server goes oom after it eats the servers 16GB slowly. normal system ram use is around 2.8GB so its not like it does not have tons of headroom.

i can use something like netdata to log and see that its indeed jelly that randomly grows. but i cant imagine why, as i said this is with no playback at all its just sitting idle. so the only things running should be the maintenance routines for the libraries like scanning and such. ive tried searching jellyfin logs but see nothing remarkable in them.

[ngosang]

I can reproduce this issue too (library with 2000 movies):

• Start Jellyfin (in my case with 1 GB memory limit in Docker)
• The app works well and the memory is ~400MB
• Go the library in the Web UI (or Android app) and scroll in the page with the movie posters. => related [Issue]: generating movie library thumbnails uses an crazy amount of memory #6877 ?
• With each page the memory goes up and never down until it reaches 1GB and it's killed by OOM.

I think you are caching poster images in the server memory or something like that. If that is true I rather prefer slow poster loading than server crash. The wise thing to do is to cache the posters and other metadata in the client.

[FilipLac]

Well, for me it stabilized when I set max RAM usable for it to 2GB and it is using somewhere around 1GB in peak. And if something happened it would just restart which will drop used ram to just 100-200MB

[gidoBOSSftw5731]

While I don't want to hijack this thread too much, I will ask, how does Jellyfin even handle caching in memory? In a perfect world, I'd want it to cache all thumbnails and all library metadata in RAM if it's available, but instead I just see the memory slowly climb (hits about 8G after a week, which is fine when the system it's in has 96, but I kind of wish I could tell it to use more.)
beta3 started about 30 minutes ago

$ sudo smem -P jellyfin -k
  PID User     Command                         Swap      USS      PSS      RSS 
2123928 root     sudo smem -P jellyfin -k           0     1.4M     2.3M     5.1M 
2123929 root     /usr/bin/python3 /usr/bin/s        0    11.1M    11.7M    13.9M 
2120303 jellyfin /usr/bin/jellyfin --webdir=        0     1.3G     1.3G     1.3G 

Music: 3879
TV: 246 Shows, 7731 Episodes
Movies: 212
Books: 3

[LP0101]

Hopping into this thread to say I'm experiencing the same issues.

Running JF in Kubernetes. On a fresh restart, it uses ~500 MiB of RAM. It will then slowly climb up, using > 5 GiB at points.

If I set a memory limit of 2GB, it looks like it mostly stabilizes around 1.5 GiB, but will still spike over 2, causing it to get OOM killed.

Screenshot before I put in memory limits, where Jellyfin's working memory continued to grow to over 5GiB:

Screenshot after putting in memory limits:

Notice that it still spikes over 2 GiB on occasion, causing it to get oomkilled, restart to 500 MiB again, then rise right back up.

[ngosang]

Still having the same problem. This graph is the memory usage in the last 6 months. The memory drops when it's killed.

First 3 months 1.5 GB memory limit. Last 3 months 2 GB limit.

With more memory the service is killed less as expected but we are wasting resources. My server has plenty of resources but some users are running in SBC or NAS.

Ideally, the program should remain stable at 500 MB, which is what it uses at startup. The rest of the memory seems to be used by caches or by resources that are not essential to play a movie. In the future, the initial amount may be reduced. I have experience with applications in .net and 500MB is much more that a service like Jellyfin should need.

[yeyeoke]

I'm experiencing the same issue.

Running Jellyfin in a LXC in Proxmox, which is limited to 4GB RAM (Previously 8, same issue then). Jellyfins RAM-usage just keeps growing over a couple of weeks which causes the LXC to hang.
Works fine for a couple of weeks after a reboot while continuing to grow to 100% RAM-usage again.

Here we can see the issue in a graph: