Jellyfin stops responding if I scan or stream #3558
Comments
|
I have the exact same Problem. I made an Account, extra to report my Problem, glad I´m not the only one. I have a nearly Identical setup. OS: Proxmox/Debian I think i didnt had that Problem in 10.5.4, but if i Downgrade with dpkg, the Problem persists. Downgrading to 10.4.x isnt working, as Jellyfin wont startup. If I scan the Library through "scheduled Tasks", it will proceed to around 92-94 Percent and then hang. Wont finish even after a Night. WebUI gets unresponsive. EDIT: I setup a completly fresh Container with Ubuntu 20.04LTS. It hangs on the Library Scan. I think it has something todo with Special characters in the Name of the Files. For Example Ö and Ä, as Jellyfin only scans till then and then hangs on "validating Media library". EDIT2: Yes i can confirm that thats the Problem. If I create a Folder with Movies that have Ä or Ö in them, Jellyfin wont scan anything. One file Shows up, but no Metadata what so ever. However, the WebUI is still responsive. EDIT3: Its not the Problem. Removed all libraries, Renamed the Files, added the Test folder again and rescanned. It broke, WebUI is unresponsive after 5 Minutes and i need to restart the Container to get it to work again. PS: I´m sorry for all the EDITs, i know this isnt a Blog or something, just tried to help, but i cant get it figured out, obviously. |
|
I have encountered same Problem. Nearly always on scanning Libraries the server hangup and Webinterface don't response anymore. |
|
Still happens in 10.6 but much less rapidly. |
|
Same problem here, my jellyfin stopped responding few days ago. My settings are quite similar to the others mentioned above: OS: Proxmox/Ubuntu 20.04 |
|
I did some "proper" testing today. First i made a fresh LXC Container, only ran apt update && apt dist-upgrade && apt autoremove. After that i thought, that the Problem may not be Jellyfin itself, but LXC instead. So i made a Full KVM with Ubuntu 20.04, did the same thing as for the container and ran a library scan. It didnt hang, stuttered, crashed or freezed in ANY way. IDK whats the Problem here, but maybe that helps someone who has more knowledge than me. |
I confirm that switching from lxc to kvm solved my issues. |
This comment has been minimized.
This comment has been minimized.
|
The people in this reddit thread : https://www.reddit.com/r/jellyfin/comments/hoki23/jellyfin_dies_when_scanning_folder/ have the same symptoms. I can confirm that using a Fedora based LXC container also works. The problem is therefore just with LXC + Debian (Ubuntu). |
I did the same thing, fixed Permissions and can confirm that it also works. I hope @aussetg can figure out whats wrong. I´m missing the knowledge todo something like that. |
|
Exact same issue with Jellyfin 10.5.5 and then the latest 10.6 with Proxmox LXC Ubuntu 18.04. Files already in the library work fine, scanning for new files freeze Jellyfin to death with the same ffprobe log as last line. Tried replacing jellyfin-ffmpeg and probe with the original ffmpeg package, it has the same effect. I would like to keep the lightweight LXC container instead of migrating to KVM if possible. |
|
I use LXC and have no such issues. Can you send me the offending file? |
|
Upgrading to 10.6 seems to have fixed it playback for me. It still doesn't want to scan my libraries. EDIT: I correct myself. After the 3rd/4th play, the server/web has become unresponsive again |
|
Same issue here... OS: Proxmox/Debian 10 Already tried a rollback of the container to the 01.07.2020, but still crashing. |
|
Same problem, I'm using jellyfin on a debian lxc container (proxmox host). Like everyone else, scanning my library is not working, after a while everything is stuck and I have to restart jellyfin, indefinitely... I can't just go for a kvm machine for I need GPU transcoding abilities, and I can't get them on kvm with my setup... |
|
I Updated Proxmox to 6.2.10 and Jellyfin still doesnt want to complete a Scan, or "quickly" starts a Episode, but at least it doesnt freeze itself to death anymore, as it seems. I only tested that one for a short time, but yeah. Update to 10.6 will fail, because it cant remove some directories as it seems.
Why cant u just use KVM? GPU Passthrough isnt that hard and works on almost any Platform. For Proxmox itself you dont need a Graphics card, as it usally runs headless. |
|
It seems to be independent from what files are new in the media folder, it just hangs forever. Permissions are fine and were fine before it started. I'm not sure if a Proxmox update or a Jellyfin update or a Ubuntu update broke it, it just stopped working once after I updated the container and the host. |
|
I'm a new user (well, I'm not now, but I was trying to be) and I'm experiencing this problem too, making it impossible to setup Jellyfin from scratch. My system: OS: Debian 10 |
For now i would just suggest to use KVM instead. You can easily migrate later to lxc if you still want. Just copy /etc/jellyfin and fix the Permissions and you good to go :) |
Doesnt seem to fix it for me. Maybe i do it wrong? |
Yes, the install is fixed but I don't get this error. The web UI loads fine but but not even the old media is playable anymore :( |
That sucks. Was really looking forward to "Syncplay" but for now it seems, that i´m forced to stay on 10.5.5 via kvm.. |
|
It's really anonoying that Jellyfin in LXC was working great and suddenly it doesn't, and even the debug log doesn't say anything what's wrong. Anyone from the Jellyfin team here? How could we debug this further? |
|
Same issue here. This started last Friday in 10.5.5 right after I renewed my Letsencrypt certificates. I don't believe that I updated the container or Proxmox before it occurred but I do know that it was working in the morning as I was tracking some newly added media. I updated to 10.6 once it was available and the UI does appear more stable but playing media or starting a scan might still cause issues with loading the UI even though the service status reports as active (running). Started a new scan this morning and it's been going for about 2 hours now (very very long). Part way through it stopped logging to /var/log/jellyfin (EDIT - this is incorrect, the jellyfin log just rotated, it was still logging correctly) but it is still logging to the syslog. OS: Ubuntu 18.04 |
Proxmox doesn't need a graphic card, but I need to transcode movies because I have a slow upload connexion at home, so I need to be able to transcode my files to lower the bitrate to watch my collection when I'm not home. I have a unused gpu and can't allow too much cpu, so the gpu decoding is my solution. I tried during days to get gpu passthrough to work, and it didn't. And yeah, I know how to do gpu passthrough, I already did some successful setups... To get back to the issue, everyone seems to be using Proxmox, is anyone using Jellyfin on LXC on another platform? as Proxmox is basically debian, it would surprise me if it worked there. |
Everybody is also using a Debian ( or Ubuntu ) LXC guest. I tried using an LXC Fedora container on Proxmox and it works flawlessly. So there's something screwy in the interaction between LXC, Debian and Jellyfin. |
So jellyfin works on fedora lxc containers? That's good to know, I'll try a fedora setup tomorrow, and archlinux too. Everyone is describing what works and what doesn't, but we still have no information on what may be causing this :/ |
I tested the installation in Docker and KVM and I don't ran into issues. |
Did you get Jellyfin back up running within a Fedora LXC? Then I would try the same. A Debian KVM is way to much overheard for my small machine. |
So, I tried with Fedora and Archlinux (and I stayed with archlinux because the From what I can see with my experience and others, the problem is only with debian/ubuntu, you should be good with fedora/archlinux... |
|
I moved Jellyfin from a Ubuntu 18.04 LXC container to a CentOS 8 (KVM) based Kubernetes cluster under the same Proxmox, it works fine. |
|
Problem seems to Persist in 10.6.1-1, even tho i can now apt-get dist-upgrade to it. |
|
Just updated my Proxmox Host and my Debian LXC with Jellyfin to version 10.6.2. Tested a few videos and they are playing fine now. Can anyone confirm this? |
|
Updated myself as well to 10.6.2, I can indeed now play videos without the whole thing crashing. Although scanning libraries still doesn't seem to work. It has found the first new item but can't seem to retrieve any data for it and hangs on the same percentage for an hour now. |
|
Having the same issue here - I think I've narrowed it down to a "deadlock" in c#, or an issue with the multithreading. I threw some additional logging functions into my build of jellyfin, and changed this to this:Repeatably, every time the task.delay is hit, the 2nd log message isn't posted, and the scan processing deadlocks. This is just the most repeatable place that I've been able to find a deadlock, it certainly happens in other places than that judging by how my jellyfin will go unresponsive at random. This is using 10.6.2, debian, proxmox hosting lxc. |
|
If I understand that right, then the Problem seems to be the TMDB Plugin ? |
|
No - I don't believe the problem is with any specific plugin, that's just one example of a repeatable place I could make at least one thread deadlock. I've since tried compiling jellyfin with native dotnet core 3.1 rather than using the standard docker build scripts and it seems to be working. It's possible that this is a bug within dotnet itself or within the docker build process (or something else entirely). I'm going to keep investigating. |
|
I have confirmed that the two builds behave completely differently - for me, building via the "native" method on debian yields a jellyfin which does not exhibit this behavior. Looking at the difference between my native build environment and the docker environment, I see that the docker pulls dotnet core 3.1.1, while my native environment uses 3.1.302. One of the changes between those two is this: dotnet/corefx#42776 which I am hypothesizing is the differentiating factor. If anyone else in this thread could try building jf 10.6.2 with dotnet 3.1.2x or above and testing for reproducibility, I think that'd be helpful to nail this down. |
Have you tried updating the docker image with a newer version? That sounds like a likely solution at the moment. |
|
Should be easily fixable - we're explicitly pulling in that older .NET release in most of the builds, but can easily bump it to 3.1.302 or later. e.g. |
|
SQLClient is the client library for Microsoft SQL Server. We use SQLite 😉 |
|
We could just remove the rate limit 😄 https://developers.themoviedb.org/3/getting-started/request-rate-limiting
|
|
I think this is the bug you're encountering dotnet/runtime#13475 |
That's definitely worth looking at removing the delay in the TMDB plugin, but as stated the delay is not specifically the problem; it just is one place I was able to repeatably identify as a deadlock. As you mentioned, SQLClient isn't used in jellyfin, but there are myriad bugfixes between dotnet sdk 3.1.100 and 3.1.302 that could be resolving the strange deadlock issues here. I'd like to lock it down to one specific commit, but the reproduction of the issue is tedious and slow so simply having a working resolution is enough for me. |
|
I Still have this issue. What I see also is that there is an zombie ffprobe process: Dotnet version is 3.1.302: any workaround without using kvm? |
|
Can confirm building from source solves this issue for me completely, it used to hang after launching ffprobe on every scan, and then just restart itself like you can see in the log, no errors just restarting. @blablup |
|
Thanks, I will compile the source. I have no problems building packages, did it a lot of times at work. |
My symptoms are EXACTLY those: linuxserver/docker-jellyfin#28
Describe the bug
If I scan libraries or transcode something Jellyfin stops responding. It's not crashed but the Web UI does't respond anymore and nothing is being written to the logs
System (please complete the following information):
To Reproduce
Go to a file, force refresh metadata. Immediately hangs
or
Go to a file, force streaming (starts very slowly somehow), skips around, will freeze eventually
Expected behavior
Scan to preceed
Logs
The text was updated successfully, but these errors were encountered: