Technical limits with Jellyfin ?

[vincen]

Hi

User since few years of Jellyfin, I'd like to share a problem I have with it. First I have to give few indications about my system.

It's running on Linux Debian 12 on a machine with a Xeon @ 2Ghz with 32 cores and 128Go of RAM. Storage is done on a ZFS array (one array is used for metadatas and one for medias storage). The ZFS arrays are able to sustain 3.5G of data transfert. Processor load is never reaching 10%.
I'm running Jellyfin server 10.8.1 and it contains 22 000 movies and around 2 500 TV shows.

I have no problem so far to play medias but the browsing and Jellyfin interface is quite sluggish and often I'm stuck on the spinning wheel for ever (whatever I use the Jellyfin Player on Linux or web browser).

Is it due at number of medias and reaching some technical limits ? or is there something that can be done ? like moving database in a MySQL system instead of flat db ?

Also something is that if Jellyfin is scanning medias, then the interface is nearly unusable till scan is over !

Thanks for advices, suqqestions,

Vincèn

[skribe]

It sounds to me as if your sqlite DB is congested. This could be caused by a few things. Share your full server log and we can take a look. There is not a way to use MySQL at this time, but sqlite should also be fine for any reasonable scale. This issue is most often caused by storage congestion in general, or plugins behaving badly. The logs are the place to start.

[bitmap]

Just a quick thing to check, I would follow skribe's advice of posting logs, but consider looking at your scheduled tasks and how long they're taking to complete and when they're scheduled.

I had issues with scheduled tasks causing occasional playback failures and network congestion during my workday and it was because Jellyfin was just absolutely blasting the TMDB API for people images. The scheduled task was set to run on a weird schedule like every 136 hours which made it fall at odd times and it took me forever to track down. I deleted that trigger and moved it to a reasonable, low-traffic timeframe and eliminated my issue. Not saying this is the same, but the symptoms are haunting echoes of the ghosts I chased for months...

[toytown]

I have about 60% of the movie collection you do, however my interface is super snappy with page load times pretty much instant.

The only exception is when JellyFin is scanning new media, in which case i also have the same issue and the interface is so sluggish it's not worth browsing. I've set this to manual so i can control at least when this happens and to be fair the scan doesn't take too long (less than a minute), my guess is that SQLite DB is the cause and the eventual move to mySql will fix this.

EDIT - Just to mention my jellyfin docker and all the metadata is on a NVMe, whilst my movies are on HDD

[vincen]

It sounds to me as if your sqlite DB is congested. This could be caused by a few things. Share your full server log and we can take a look. There is not a way to use MySQL at this time, but sqlite should also be fine for any reasonable scale. This issue is most often caused by storage congestion in general, or plugins behaving badly. The logs are the place to start.

Sorry for late answers but as notifications by email are not working I saw them late  Regarding logs can I share them without precautions ? nothing sensible in it ? I have seen there is the IP address of server that I can remove easily, anything else before I post it here ?

Just a quick thing to check, I would follow skribe's advice of posting logs, but consider looking at your scheduled tasks and how long they're taking to complete and when they're scheduled.
I had issues with scheduled tasks causing occasional playback failures and network congestion during my workday and it was because Jellyfin was just absolutely blasting the TMDB API for people images. The scheduled task was set to run on a weird schedule like every 136 hours which made it fall at odd times and it took me forever to track down. I deleted that trigger and moved it to a reasonable, low-traffic timeframe and eliminated my issue. Not saying this is the same, but the symptoms are haunting echoes of the ghosts I chased for months...

Thanks for advice about scheduled tasks and yeah I had quite a few setup like "Run every xxxx hours" so I have rescheduled all of them to run by night on different days of week and with time limit to be sure they are always running during quiet times

I have about 60% of the movie collection you do, however my interface is super snappy with page load times pretty much instant. 
The only exception is when JellyFin is scanning new media, in which case i also have the same issue and the interface is so sluggish it's not worth browsing.  I've set this to manual so i can control at least when this happens and to be fair the scan doesn't take too long (less than a minute), my guess is that SQLite DB is the cause and the eventual move to mySql will fix this.
EDIT - Just to mention my jellyfin docker and all the metadata is on a NVMe, whilst my movies are on HDD

I have fully deactivated the automatic scan of libraries because I had already noticed same problem you point out: when a scan is running the server is nearly down in term of use
Storage should not be a problem as metadata folder is on a dedicated ZFS array and an other one for media files and both have around 2.5g of data transfer rate sustained.

[TheDreadPirate]

Sorry for late answers but as notifications by email are not working I saw them late  Regarding logs can I share them without precautions ? nothing sensible in it ? I have seen there is the IP address of server that I can remove easily, anything else before I post it here ?

You should remove any WAN IPs from your logs and user names.

Then upload your log to sourcebin or pastebin and share the link here.

[vincen]

You should remove any WAN IPs from your logs and user names.

Thanks here is the file: https://group.vincen.org/index.php/s/ktjqTgjNKQ8Spzd

[TheDreadPirate]

Theres a lot going on in that log. The only thing out of the ordinary is ffmpeg crashing a bunch. Can you share those ffmpeg logs with us?

Is the database on the ZFS array? Located in /var/lib/jellyfin/data/. Or is that directory on an SSD?

[vincen]

Theres a lot going on in that log.  The only thing out of the ordinary is ffmpeg crashing a bunch.  Can you share those ffmpeg logs with us?

For ffmpeg I have a bunch of log files as it generates one for each encoding request  Here are last ones, let me know if you need anything else  For ffmpeg I have installed on system the special version for jellyfin as indicated in configuration documentation ! https://group.vincen.org/index.php/s/MZXDYq4NxAjYRZs

Is the database on the ZFS array?  Located in /var/lib/jellyfin/data/.  Or is that directory on an SSD?

The database is on the zfs array. I have done a symlimk of /var/lib/jellyfin/data

I have noticed quite some improvements already by resetup of scheduled tasks so they run only during quiet time as suggested by @bitmap

[TheDreadPirate]

Having your database on a ZFS file system is possible, it is very much not recommended. Database READ performance is fine, but when you are writing to a database that is on a ZFS file system there are serious performance penalties.

And because your database is big, that adds to the performance penalty.

If possible, move your database to an EXT4 file system.