SOLVED: New TV, unable to skip chapter without a memory leak

[podonnell]

Just set up a new LG G3 TV. Testing a particular title, I am having some type of memory issue. I attempt to skip ahead a single chapter on a title, and the TV hangs for about 2 minutes and then says "this app will now restart to free up more memory"
I'm very perplexed, as I have an LG C1 that does not experience this issue, using the same title, connected to the network the same way.
I can also do this on my PC, and the chapter skip takes just seconds.

Here is the ffmpeg log which receives a q code from my G3 after about 2 minutes:
https://pastebin.com/4wFRZG1p

Here is the section of my jellyfin log where I think I see the issue happening:
[2024-07-04 20:41:08.390 -04:00] [INF] [29] MediaBrowser.Controller.MediaEncoding.TranscodingJob: Stopping ffmpeg process with q command for "\Jellyfin\Data\cache\transcodes\db1581c647ddfadc7fb592dac996bf8e.m3u8"
[2024-07-04 20:41:08.597 -04:00] [INF] [29] MediaBrowser.MediaEncoding.Transcoding.TranscodeManager: FFmpeg exited with code 0
[2024-07-04 20:41:08.602 -04:00] [INF] [29] MediaBrowser.MediaEncoding.Transcoding.TranscodeManager: "ffmpeg" "-analyzeduration 200M -probesize 1G -ss 00:11:43.368 -noaccurate_seek -fflags +genpts  -i file:\"S:\CIPHER\Movies\Blade Runner 2049 (2017)\Blade Runner 2049 (2017) [imdbid-tt1856101] - [ATMOS Hybrid REMUX][Remux-2160p][DV HDR10][TrueHD Atmos 7.1][HEVC]-FraMeSToR.mkv\" -map_metadata -1 -map_chapters -1 -threads 0 -map 0:0 -map 0:2 -map -0:s -codec:v:0 copy -tag:v:0 dvh1 -strict -2 -bsf:v hevc_mp4toannexb -start_at_zero -codec:a:0 copy -copyts -avoid_negative_ts disabled -max_muxing_queue_size 2048 -f hls -max_delay 5000000 -hls_time 6 -hls_segment_type fmp4 -hls_fmp4_init_filename \"\Jellyfin\Data\cache\transcodes\db1581c647ddfadc7fb592dac996bf8e-1.mp4\" -start_number 117 -hls_segment_filename \"\Jellyfin\Data\cache\transcodes\db1581c647ddfadc7fb592dac996bf8e%d.mp4\" -hls_playlist_type vod -hls_list_size 0 -y \"\Jellyfin\Data\cache\transcodes\db1581c647ddfadc7fb592dac996bf8e.m3u8\""

My server is older hardware, 4790k CPU with VAAPI transcoding. But again, this works perfectly fine on my LG C1 which is about 3 years older.

This TV is doing a REMUX due to file container in one test, and in another test I enabled TrueHD and it was doing an audio transcode, but the results are the same either way.

Any ideas what could be happening?

[podonnell]

I've had two new instances of this in the past 48 hours on the new TV.

https://pastebin.com/Yk0RdUhQ

This one was from another Jellyfin server entirely, and one that is much more powerful than my normal one.
This instance was a bit different as there was never a memory error message that came up, but instead I tried to seek in to the video and it took 2-3 minutes before the video just closed, and it never seeked forward.

The second instance was from my normal server, and I was trying to resume a video that I had stopped in the middle of the night before. It would not resume at all, and the screen stayed black for a minute. At one point a subtitle appeared if I recall correctly, but it didn't seem like the video was actually playing.

I tried to resume the video a few times, but it would stay a black screen and never 'seek ahead' to this point.
[2024-07-07 12:37:05.138 -04:00] [INF] [12] MediaBrowser.Controller.MediaEncoding.TranscodingJob: Stopping ffmpeg process with q command
These seem to be the same errors in the log around that time.

I eventually did a 'start from beginning' and the video played just fine. Although I was once again unable to 'skip chapter', the video would freeze and never jump ahead.

What could be causing these issues which seemingly all have to do with seeking ahead in certain videos?

[TheDreadPirate]

That's the ffmpeg log, which looks normal. Can you share the main server log?

[podonnell]

That's the ffmpeg log, which looks normal.  Can you share the main server log?

Just tried on both my servers to see if it generated anything useful.

This one was a 30 second fast forward that took about 3 minutes to complete the fast forward:

https://pastebin.com/SCWpyfck

And this one was a chapter skip right after starting a title, waited 4 minutes and backed out. It didn't change a single frame after pressing the chapter skip, just stayed frozen:

https://pastebin.com/UneiMivz

The timeout on port 5000 is known as that is a webhook I don't currently have running, and Jellyfin plugin disable button does not work at all.

[TheDreadPirate]

On your TV, can you go to the playback settings in the Jellyfin app and uncheck "prefer fMP4"?

[podonnell]

On your TV, can you go to the playback settings in the Jellyfin app and uncheck "prefer fMP4"?

This did fix the issue. I can skip ahead now.
However, it also prevented the video from playing in HDR and requires a transcode on the video now.

[TheDreadPirate]

Yeah. There is an issue in the WebOS app (not sure if its us or WebOS) that causes issues when we send fMP4. But fMP4 is required by WebOS for better HDR compatibility.

This is a known issue and is being worked on.

https://github.com/jellyfin/jellyfin-webos/issues/236

The related fix.

https://github.com/jellyfin/jellyfin-web/pull/5553

[podonnell]

Yeah.  There is an issue in the WebOS app (not sure if its us or WebOS) that causes issues when we send fMP4.  But fMP4 is required by WebOS for better HDR compatibility.

This is a known issue and is being worked on.

https://github.com/jellyfin/jellyfin-webos/issues/236

The related fix.

https://github.com/jellyfin/jellyfin-web/pull/5553

Awaiting this fix, and have turned off the fMP4 for the time being.

That said, I did have an odd issue that I have only experienced on a title that I had to turn off fMP4 for. The video would stop playing on its own and not go beyond a certain point. Only until I quit the application and restarted the video would it go beyond. It was a container remux only. The ffmpeg logs showed nothing, but I did find this in the main log around the time it happened. Unsure if related:

[2024-07-15 01:20:48.560 -04:00] [ERR] [9] Microsoft.AspNetCore.Server.Kestrel: Unexpected exception in TimingPipeFlusher.FlushAsync.
System.Net.Sockets.SocketException (10022): An invalid argument was supplied.
  at System.IO.Pipelines.Pipe.GetFlushAsyncResult()
  at Microsoft.AspNetCore.Server.Kestrel.Core.Internal.Infrastructure.PipeWriterHelpers.ConcurrentPipeWriter.FlushAsyncAwaited(ValueTask1 flushTask, CancellationToken cancellationToken)
  at Microsoft.AspNetCore.Server.Kestrel.Core.Internal.Infrastructure.PipeWriterHelpers.TimingPipeFlusher.TimeFlushAsyncAwaited(ValueTask1 pipeFlushTask, MinDataRate minRate, IHttpOutputAborter outputAborter, CancellationToken cancellationToken)

[TheDreadPirate]

Does this sound like your problem maybe? Regarding Jellyfin stopping playback.

https://github.com/jellyfin/jellyfin/issues/12182

[podonnell]

Does this sound like your problem maybe?  Regarding Jellyfin stopping playback.

https://github.com/jellyfin/jellyfin/issues/12182

Negative, no docker for me. Also this has never happened on another source before. First instance was on a track that I wouldn't have played before due to the fMP4 causing me to be unable to pause or restart. Since I turned that off, I was able to watch this video, but the pausing happened randomly. I will keep an eye out for other instances of this happening. Would 'HLS' videos be the type that are affected by this setting only?