transcode, exit code: 1 #39
Comments
|
I've actually notice this same issue and haven't been able to track down the reason why the task is terminated. My guess is that the transcoding process gets terminated as part of regular operations and it just takes longer than what plex expects to start up again. I really should verify this hypothesis, but just haven't had a chance to do that yet. |
|
Ya, I wasn't able to track it down either in a day of troubleshooting. Kept switching between enabling kube-plex and disabling it with the two videos failing with in my plex web app & mobile web app, but the videos worked with direct play and htpc. Eventually, at end of day, I put in an issue at a plex forum and after the first response I went to gather up some logs and everything started working with kube-plex disabled. I'm not sure if that's a clue, that two videos were failing in both the plex web app and plex mobile with kube-plex disabled, as well as with kube-plex and the pod exiting with code 1... but then, after several hours, the plex web app and plex mobile started working with kube-plex disabled, and now its just with kube-plex enabled that it doesn't work. |
|
So, I made some headway on this. First, I had to change the loglevel manually (the loglevel attribute is broken, I had a PR I'll put up for that). Secondly, I shimmed the transcode-launcher with a script that sleeps on fail. Otherwise it's impossible to debug. There is some output from the Plex transcoder about an invalid argument, but I haven't dug into that side too much yet. Sorry for the long paste: |
|
Switched to kube-plex-kube yesterday, noticed it wasn't playing some formats and switched back, haven't looked at this in awhile. Does the kube-plex strategy not support the same formats as the plex image? |
|
It did with some exceptions, but Plex seems to have changed something in the transcoding process so that it's more dependend on the main process. I noticed that formats that used to be supported now require transcoding without kube-plex. (see note in readme) I didn't look too much into it, but I'm assuming it's doing the same thing as with EAE where it writes a bunch of files to |
|
Did u happen to post on the plex forums or otherwise try to contact the developers? Maybe they did the change on purpose to break this spin-off? What did u switch to? Perhaps I could try something new. |
|
Could /tmp be a rwx shared by plex and transcoder pods? I know, I obviously don't really know how kube-plex works... it was just a cool project. |
|
Well, I gave it a shot at the plex forum: |
|
I haven't asked the devs, but I have very low confidence that they would agree to help. We are dealing with such low level details that they most likely don't want anyone to mess with them. If they did, then they would need to support the current implementation and would be restricted from iterating on their product. Not to mention that the Plex software is rapidly moving away from self hosted libraries with a clear focus on their own cloud services. And honestly I fully understand and appreciate that move, even if it makes our lives more painful. I've moved to using Jellyfin. The main driver for the move has been Trakt and Anime support. Or the plugin support in general. As for the sharing of |
|
Didn't mean to close it, will keep it open, who knows, a dev might have a moment and decide to tweak things just a bit in the direction we need. |
|
Maybe we could make a kube-jellyfin given that jellyfin is more open. I was directed to rffmpeg as a starting point. Jellyfin roadmap: https://jellyfin.org/posts/release-roadmap-10.10.0/ The author there, Joshua Boniface (Project Leader), is on this thread at rffmpeg: |
|
If I may comment here? The EAE, which is a program provided to Plex by Dolby, requires:
Examination of the Plex logs will show when the inotify fails (often because the inotify table is already full) as well as show if unable to obtain the exclusive file lock. PS: Plex is not rapidly moving away from hosted media. It is at Plex's core. Folks have seen more visible activity on the streaming side because of the amount of work quietly being done on PMS internals. (new transcoder development for one thing) |
Thanks for clarifying, that lines up with what I remember from my past investigation. And sadly confirms that there really aren't any simple paths forward.
Yeah, I think that came out wrong. I didn't mean to imply that the hosted media is going away, but that the features that many of us have grown to rely are being phased out and streaming is being pushed pretty heavily from the users perspective. And like I said, I fully understand this direction from both business and product viability perspectives. I would even say that it's the right move to make. But for me that has meant that losing plugin support, which was a critical part of my Plex experience has meant that it's time to look elsewhere. And in my opinion, that is OK too. I'll still keep following the progress and might come back some day. |
|
Regarding Plug-ins. The Python-based plug-ins deprecation and ultimate support removal was announced over 3 years ago well before the new C++ Agents (scanner + metadata) even began their development. Since that time, the official API (https / curl-based) has been improved and better documented. If you look at two of the most common tools formerly as plug-ins, Tautulli and WebTools-NG exclusively use the API without any loss of function. Yes, it's a pain to rewrite plug-ins in C/C++ and use libcurl versus Python but it's really not that difficult. You now have the freedom to use other compiled programming languages as well. I see it as a mixed bag but, given what's been gained, (I can do a complete rebuild in 30 minutes versus 7+ hours ) I think we've gained more than we've lost. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
With some newer videos I've been experiencing the transcoding failing, while everything works if I disable kube-plex. Since the transcoding pod is just using the libraries from the pms image, can we say that it must be crashing due to resources such as memory / cpu? I'm not seeing any cpu or memory errors if I 'k describe pod pms-...'. Here's from the description of a pms pod that errored out:
The text was updated successfully, but these errors were encountered: