SOLVED: Jellyfin crashes randomly

[Esmail EL BoB]

Hello, I noticed recently in last couple of months that Jellyfin is unstable with me and it crashes a lot even tho I used to host 10.9.x and it was fine so I'm not sure if something change on my part of what exactly but all I can say for sure is jellyfin started to crash a lot more lately and in random times someday without a crash and other days 2 caches in same day.

I use docker compos on debian and jellyfin 10.10.5 and when that happen only jellyfin crashes not other containers or server itself, it's only jellyfin becomes frozen for sometime then unfreeze itself. I tried to look up logs multiple times but no luck so I hope you can help me pinpoint the issue

Logs: https://privatebin.net/?2b24c6c3524c0f98...s4ENX8vo5u

Thanks in advance

SOLVED: Re-install seems fixed it

[TheDreadPirate]

What are the full specs of your server? And how many users do you typically have going at the same time?

It looks like you have a lot going on and it simply ran out of resources.

[Esmail EL BoB]

Hello, The hardware is: CPU: AMD Ryzen 7 3700X, RAM: 2x RAM 32768 MB DDR4 ECC and HDD2x SSD M.2 NVMe 1 TB. For users, highest i got like 8 streams at same time (not transoding) but I remember it did not freeze or had any issues.

[TheDreadPirate]

Ok. I missed that you disabled transcoding. I see that in the logs now.

I see the subbuzz plugin running a lot. What is the CPU usage like when that runs? I've never used it myself. It looks like it does some OCR conversion maybe?

I also see these generic errors pretty regularly.

Code:

[2025-02-10 18:35:37.258 +00:00] [ERR] [18] Namo.Plugin.InPlayerEpisodePreview.Api.InPlayerPreviewController: InPlayerEpisodePreviewPlugin: "Namo.Plugin.InPlayerEpisodePreview.Web.InPlayerPreview.js"

And I saw this message after one of the restarts. Slightly concerning that this cleanup job would fail. Kind of adds to my statement earlier, and the log messages, regarding some sort of resource starvation.

Code:

[2025-02-10 18:29:50.680 +00:00] [ERR] [19] Emby.Server.Implementations.ScheduledTasks.Tasks.DeleteCacheFileTask: Error deleting directory "/cache/transcodes"
System.IO.IOException: Device or resource busy : '/cache/transcodes'
   at System.IO.FileSystem.RemoveEmptyDirectory(String fullPath, Boolean topLevel, Boolean throwWhenNotEmpty)
   at System.IO.Directory.Delete(String path, Boolean recursive)
   at MediaBrowser.Controller.IO.FileSystemHelper.DeleteEmptyFolders(IFileSystem fileSystem, String path, ILogger logger)

Might not be a hardware resource that is being starved. Possibly some sort of OS or Jellyfin internal queue.

[Esmail EL BoB]

Hello,
I first switched from opensutbitles to subbuzz because i thought it's the issue, and here is a random htop: https://ibb.co/Zzbkk4mQ so I would say subbuzz is not that CPU intensive.

I agree it can be jellyfin or OS issue as well, but how to investigate more? and I'm about to re-install the server again but I backed up jellyfin data to not start from zero again so I hope it's more of OS issue because I really don't want to lose JF data

So I will come back again to post results and I hope if it's jellyfin we can find a way without purge JF data

Thank you so much for taking from your time and helping.

[TheDreadPirate]

I'm not entirely sure what the problem could be. Nothing in the log is a smoking gun.

Before you try to reinstall the OS, can you share your docker compose?

[Esmail EL BoB]

I'm sorry I don't get notifications when I get reply on the post and here’s my Docker Compose setup: https://privatebin.net/?c514fb5ae9f9c5f5...Heho4rSEHr. I’m using mods, which is why some volumes are mapped. I tried disabling them, but the issue used to persist.

I reinstalled the OS, but even after that, Jellyfin froze again. I’m not sure if it was due to a library scan and backup (file uploads) or a broken Jellyfin setup. I’ll grab the logs the next time it happens since the current logs are too long (it ran for over a day before freezing). I didn’t want to overwhelm you with so much noise in logs, so I’ll wait for the next freeze and upload it (hopefully, it won’t happen but if happens).

[TheDreadPirate]

What is the goal of changing the ffmpeg probe size and duration? I don't think what you've set is the cause of your problem. Just curious.

Some unrelated comments.

PUID and PGID is not used with the official image. The "user: 1000:1000" is enough.

This line is unnecessary.

Code:

- ./jellyfin/cache/transcodes:/cache/transcodes

As is this line.

Code:

- JELLYFIN_WEB_DIR=/jellyfin/jellyfin-web

If your reverse proxy is running on the same host as Jellyfin, which I suspect is the case based on your port publishing to localhost only, you can instead use "expose" instead of "ports".

Code:

expose:
      - 8096

[Esmail EL BoB]

I set custom ffmpeg probe size and duration because when I started to host jellyfin a friend who used to host jellyfin before me recommended me these as better settings -- But I took all of your notes, I commented out the unnecessary lines such as /cache/transcode and will test it and come back to you.
Thank you

[Esmail EL BoB]

It happened again, so I'm sure it's jellyfin files at this point. but I want to make sure before I nuke jellyfin. Here are the logs: https://send.vis.ee/download/a5d3ab8ab92...EqEsMDCnhQ (It's 17th and 18th since I was not sure due to timezone difference between me and server) and I know I said next time hopefully it will be less noisy but I could not catch it