CPU

[TheDreadPirate]

Go to the path you mounted /config for the jellyfin container. There will be a log folder with a log named log_20241017.log. Copy and paste the contents of that into pastebin.com.

[thbkk]

Hello, here is the protocol. This time it happened while watching a movie. The CPU spike and the delay of the movie started at 11pm (23:00).
It could be that the fast forward option of the movie was used around that time. Actually, it is a direct play. So I thought the fast forward shouldn't have any effect.

Here the logs: Log

The movie was running via jellyfin but with external Player VLC

[TheDreadPirate]

Jellyfin is just sending the file. Not transcoding or remuxing. No trickplay or chapter image extraction. The only thing I see is the webhook notifications. Try disabling that plugin and see if that changes anything.

[thbkk]

Jellyfin is just sending the file.  Not transcoding or remuxing.  No trickplay or chapter image extraction.  The only thing I see is the webhook notifications.  Try disabling that plugin and see if that changes anything.

Look row 19-27. There are warnings and this matches with the CPU PEAKS at this time

[TheDreadPirate]

I see that, but I don't see how sending a file unmodified would do that. The only thing that makes sense is if you have other stuff going on that would consume memory bandwidth or network bandwidth or disk IO.

Without transcoding, but the CPU is still peaking, indicates that some data path is saturated and the CPU is struggling to move data around fast enough.

[thbkk]

Ok, let me investigate it. But the peaks only appear when using jellyfin. I will switch the arr stack to a mini server with more power.

What do you recommend for Jellyfin. Jellyfin in a separated LXC proxmox or all in one VM with docker container?

[TheDreadPirate]

Is the current docker instance in a VM? Or Docker on the host OS?

Order of preference.

1) Direct install on bare metal OS
2) Docker install on bare metal OS
3) Direct install in LXC

And a distant last

4) Docker install in VM

With the limited performance of the RPi4 you really need to minimize overhead that virtualization causes. Docker and LXCs minimize a lot of that since they use the host kernel while VMs use their own kernel with virtualized hardware, introducing a lot of overhead.

[thbkk]

Is the current docker instance in a VM?  Or Docker on the host OS?

Order of preference.

1) Direct install on bare metal OS
2) Docker install on bare metal OS
3) Direct install in LXC

And a distant last

4) Docker install in VM

With the limited performance of the RPi4 you really need to minimize overhead that virtualization causes.  Docker and LXCs minimize a lot of that since they use the host kernel while VMs use their own kernel with virtualized hardware, introducing a lot of overhead.

Thanks, but due to the low performance, I bought a mini server (Lenovo ThinkCentre M720q Tiny - Intel i5 8400t). My plan is to use a VM with Docker containers for the ARR stack, hoping that this will solve the issues. A friend has the same mini server but experienced occasional freezing when using fast forward. He switched to Plex and hasn’t had any issues since. Any ideas?

[TheDreadPirate]

How long of a "freeze" are we talking? Without more info it is hard to say. Maybe transcoding?

[thbkk]

How long of a "freeze" are we talking?  Without more info it is hard to say.  Maybe transcoding?

I will concentrate on the new setup. 
You said lxc is better than VM. How difficult is the Passthrough or not needed?
My plan is jellyfin in lxc with Passthrough and the rest of arr in one Devian VM.
Make that sense?