BUG : Stopping a stream does not kill all processes

Plex Media Server Desktops & Laptops
,
1

Server Version#:

1.16.0.1226 (Docker / Debian)

Player Version#:

Version 3.104.1

Hi everyone,

Just experimented HW transcoding, and a i note a BIG bug.

When a stream is transcoding, we’ve got this :

STREAM

and processes :

fuser

But when the stream is stopped, or canceled :slight_smile:

after

The PID 21771 is still alive, it’s a relicate of the streaming, et it keep my GPU on work ! (P0 state).

Must kill this process at the transcode killing signal … because it keep active the transcoding work

AFTER-SMI

You can see the P0 state, 21w and 0 transcoding process, if i kill the 21771 :

AFTER-SMI-KILL

EDIT : I found the bug !

There are 2 process pid, “Plex Media Serv” (PID 001) and his child “Plex Transcoder” (PID 002).

Stop or cancelled streaming kill the child, but not the parent !

You must kill the parent ( which will stop the child process ), or just assign PDI 002 to nvidia

1 Like
2

You created the bug. Your transcoder script to force Decode/Encode (not officially supported by Plex) with Nvidia isn’t doing any cleanup.

3

Remove the patch (recreate from scratch with docker), bug is still here :slight_smile:

PLEX-WITHOYUT-PATCH
PLEX-WITHOYUT-PATCH701×331 68.4 KB

FUSE-WITHOUT

And after kill the stream

WITHOUT-AFTER SMI-WITHOUT

Note the 19mo (but only 6 for Xorg, 13 mo for “hidden” process “Plex media serv”)

It seems that Plex attributes a process (parent of transcoding in fact) that is not transcoding to the GPU

4

@anon18523487 Here is my topic :wink:

5

So is this just a memory/process issue? I see chuck has been answering in that other thread. He is more knowledgeable than me on that stuff.

If you were having quality/transcode issues, that’s something I could help look into.

Does this issue have an affect on streaming, other than the extra memory/cpu usage?

6

It’s not a quality issue :wink:

To put it in a nutshell, killing or cancel transcoding streamdoes not release the GPU, which results in an unnecessary occupation (because of a parent process that does not concern transcoding) (about 3 times more than idle) of the GPU.

This one consumes more, and heats more

7

Ok, that is beyond my knowledge. If you don’t hear back from chuck after a while, let me know and I’ll see if there is someone else that can help.

8

Ok thanks, just for your information , all the peoples who test in Reddit or in circle of friends have the issue.

I will directly blame nvidia, but all other competing solutions do not have the bug, as well as native ffmpeg … so … it’s probably plex ( or Plex / Nvidia interaction)

9

If this helps, I was testing how many transcodes I could do on my laptop with the server on windows 10 with Intel i7-3610QM and Nvidia GT 650M. I tested with multiple Apple devices. Tested as follows, streamed direct play on 6 devices 1 by 1. Then began changing the quality so it would transcode 1 by 1 to find my limit. The laptop was fine with 2 streams showing (hw) twice under video on the dashboard. I also set my CPU not to exceed 50% on Power Saver mode, 70% at Balanced mode and up to 100% in Performance mode. After starting a 3rd video transcoding with it in Power Saver mode, it would only show (hw) once under video in the dashboard vs 2 of them on the first 2 transcodes. The laptop would then overheat hitting high 80’s Celsius and the CPU usage was exceeding the 50% limit for Power Saver mode to 80+% CPU usage to where I had to stop the videos.

10

If your CPU is hitting > 80C you’ve got a thermal issue somewhere. With a proper setup you should not be exceeding 75C even at 100% load.

11

It is an old laptop and does meet my needs and was only testing the transcode ability. I only direct play but wanted to know in case a user transcodes a stream. So I just limit to to 2 transcodes so it should prevent it from overheating.

12

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.