Some shows suddenly take minutes loading playback

[Marco2G]

Hi everyone
I have noticed two shows that are stuck on the loading screen for a literal minute or more before they start playing.
Other shows seem fine. And the first few seasons of one of the shows in question worked normally.
Can anyone give me a hint where I start troubleshooting this?

[TheDreadPirate]

Can you share your logs with us via sourcebin or pastebin. Sounds like you are transcoding. Also mention which show you are having trouble with to make sure we look in the right place in the log.

[Marco2G]

Well, I hope I'm doing it right. I sanitized the jellyfin log of my TLD.

So the show in question here is Clarkson's Farm. I added one for Little House on the Prairie, which does not have the issue.

I'll try and find a Big Bang Theory one, which has the issue too, from another day.

https://pastebin.com/0drPm2bJ
https://pastebin.com/ANhq4VTd
https://pastebin.com/Af4qUkVG
https://pastebin.com/sJdCUvu4
https://pastebin.com/UA1RM2h5
https://pastebin.com/buSw0tTn
https://pastebin.com/wG6A4NRc

[TheDreadPirate]

What client are you trying to play on? It says it is transcoding due to "Directplay Error". Not clear why it takes more than a minute to even start the transcode.

Also, server specs? Which Linux distro?

[Marco2G]

This is an Ubuntu 22.04 VM with two CPU cores of E5-2650 v2 @ 2.60GHz, 4GB RAM and a Quadro P400 for hardware acceleration.

I have this issue both on two independent computers running the web client in Firefox and on my Android through the Android client. It's the same on both.

Edit: Just tried T2 in 4k. Started inside 5 seconds. So it isn't a question of file size or network latency.

Which brings me to describing my setup a bit more. I have a frankenserver build that runs ESXi on which I am running TrueNAS in a VM. I have PCI passthroughed an HBA so the trueNAS can work the 12ish spindles in ZFS. I also have three NVMe gumsticks in another ZFS pool. The Jellyfin VM runs on that. The Media is mounted into Jellyfin via NFS. Since both it and the TrueNAS VM run on the same hypervisor, I'd say network traffic should remain on the virtual switch. Both for the Jellyfin's iSCSI NVMe datastore and the NFS mount for the media. If it doesn't, the Machine is connected to the physical switch via 10G fiber.

[TheDreadPirate]

Nothing about your setup would cause this issue, AFAICT. You can try turning on debug logging in Jellyfin, though I have rarely found it helpful.

https://jellyfin.org/docs/general/admini...ug-logging

[Marco2G]

That last part would be something the devs might want to have a look into

I have set the new log level and from my understanding, that does not take a restart. Hope the change took hold.

https://pastebin.com/e6mmQtEz
https://pastebin.com/DHLziQWV

I timed it this time and it was about 90 seconds. I mean don't get me wrong, this is a very survivable first world problem... and once the thing has started playing for more than a few seconds, you can pause and start it normally. It's just if it was something obvious I could fix, I would prefer to.

[TheDreadPirate]

Long shot. Can you check that the time on your server and clients are in sync?

[Marco2G]

Well, the timezone was wrong but the time itself is spot-on.

[TheDreadPirate]

Can you try enabling transcode throttling? Dashboard > Playback, all the way at the bottom.