SOLVED: Jellyfin crashes unraid while media scanning

[grobalt]

Hi all,

i am figuring out why Jellyfin crashes my whole server - makes it unresponsive - without any useful logfile entry of rsyslog or jellyfin container. After a while i figured out that it may be an out of memory problem.

Hardware is a 13600k with 16GB memory, so decent amount of cores and usually ram. Since i limited the ram of the docker container to 10G and added a swapfile of 32G it seems to run stable, but what is happening ?

top - 16:55:38 up  1:06,  0 users,  load average: 261.72, 265.79, 230.91
Tasks: 272 total,  2 running, 270 sleeping,  0 stopped,  0 zombie
%Cpu(s):  0.7 us,  5.7 sy,  0.0 ni, 47.6 id, 45.9 wa,  0.0 hi,  0.1 si,  0.0 st
MiB Mem :  15829.8 total,    168.7 free,  11559.1 used,  4101.9 buff/cache
MiB Swap:  32768.0 total,      0.0 free,  32768.0 used.  3732.1 avail Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM    TIME+ COMMAND                                                                                                       
7637 abc      20  0  40.3g  20608  15564 S  2.7  0.1  0:00.66 jellyfin                                                                                                     
5944 abc      20  0  302388  40612    680 D  1.3  0.3  0:02.05 ffprobe                                                                                                       
5100 abc      20  0  330920  41444    588 D  0.7  0.3  0:02.03 ffprobe                                                                                                       
5199 abc      20  0  302048  37844    744 D  0.7  0.2  0:01.86 ffprobe                                                                                                       
5200 abc      20  0  290916  43600    928 D  0.7  0.3  0:02.02 ffprobe                                                                                                       
5254 abc      20  0  295832  32060    980 D  0.7  0.2  0:01.81 ffprobe                                                                                                       
5607 abc      20  0  312240  34448    692 D  0.7  0.2  0:01.84 ffprobe                                                                                                       
5666 abc      20  0  335000  44520    680 D  0.7  0.3  0:01.90 ffprobe                                                                                                       
5849 abc      20  0  330568  39880    684 D  0.7  0.2  0:02.29 ffprobe                                                                                                       
5863 abc      20  0  330984  45116    932 D  0.7  0.3  0:01.87 ffprobe                                                                                                       
5998 abc      20  0  287964  32768    940 D  0.7  0.2  0:01.99 ffprobe                                                                                                       
6004 abc      20  0  293196  35648    656 D  0.7  0.2  0:01.97 ffprobe                                                                                                       
6125 abc      20  0  334868  43864    996 D  0.7  0.3  0:02.07 ffprobe     


Jellyfin launches hundreds of ffprobes and ... most likely that is the reason of the freeze. I will see what happens when it eat up the 32GB swap file as well ... Any hint what to change ? i disabled all features in Jellyfin, it is a naked installation with no trickplay, no chapter pictures etc.

[grobalt]

And crashed after around 25 minutes ...

[TheDreadPirate]

Are you using Docker to run Jellyfin? Which image? Linuxserver? Or the official image?

The ffprobe's are probably from a library scan. You can limit the number that spawns in Dashboard > General > Parallel library scan tasks limit.

[grobalt]

now i changed JELLYFIN_FFmpeg__analyzeduration to 3M ... run succesful, all 1600 movies scanned but a lot of no codecs found. now running again with 10M and will report what changed

---

Are you using Docker to run Jellyfin?  Which image?  Linuxserver?  Or the official image?

The ffprobe's are probably from a library scan.  You can limit the number that spawns in Dashboard > General > Parallel library scan tasks limit.

i tried with official image as well as linuxserver.io (current) - both behaved same.

[grobalt]

JELLYFIN_FFmpeg__analyzeduration=3M ... rushing through all movies in a few minutes but issues with some codecs. Tried 10M ... better ... but still swapping 20G additional to 12G for the container .. ridiculous.

Now limited the scans to 4 max parallel and JELLYFIN_FFmpeg__analyzeduration=300M default

*still running* ... i think this "endless" parallel scans will break most disk based systems ... if i googled correctly, this was introduced with 10.9

[grobalt]

https://github.com/jellyfin/jellyfin/issues/12203
issue raised .. default setting is bad