Jellyfin in Synology docker has a very high CPU usage during HWA

[wzl778633]

Hi! I got a DS720+ to run Jellyfin on it. CPU is Intel Celeron J4125 so I believe it has a GPU inside. I upgraded my jellyfin recently to 10.10.3 and I mapped the devices for hardware acceleration in docker compose file, everything should be prepared and I enabled QSV as my way to do HWA. The video can "start" to play but the transcode speed is just horrible (basically the speed that cannot play) and the CPU usage is always closed to 100% during transcoding. I expected 30%~ CPU usage if the transcoding is using GPU correctly instead of using CPU. Any help for this？ I checked the logs from Jellyfin and I don't see any errors. I will put jellyfin-ffmpeg's logs here. 

I tried QSV and VVAPI, non of them are working "correctly". They are all landed in same situation.

Jellyfin: 10.10.3
Device:  Synology DS720+, DSM 7
environment: Docker

[TheDreadPirate]

What client are you using that would request a VP9 transcode? I didn't even know Jellyfin would allow that.

VP9 DECODING is supported by the iGPU but, somehow, you've gotten Jellyfin to ENCODE to VP9 which your iGPU does not support doing.

[wzl778633]

What client are you using that would request a VP9 transcode?  I didn't even know Jellyfin would allow that.

VP9 DECODING is supported by the iGPU but, somehow, you've gotten Jellyfin to ENCODE to VP9 which your iGPU does not support doing.

Thank for the reply! I'm using chromium to "test" the transcoding process. I think this browser should support h264 but I don't know why it is trying to encode to VP9...
This is the setting I'm using for:

[TheDreadPirate]

Can you share your full Jellyfin log via privatebin.net?

[TheDreadPirate]

Can you check if Jellyfin has permission to write to /cache/transcodes? That is the default transcode directory, but yours is writing to /config/transcodes. I'm wondering if this weird behavior is due to not being able to write to the actual transcode directory.

[wzl778633]

https://privatebin.net/?bb433f791d5a9a51...fdUdLmSwu6 This is the logs from jellyfin
https://privatebin.net/?87dbb88e1526bb8e...RSL7tBTrTW This is the logs from jellyfin-ffmpeg
I don't think writing to /config/transcodes is the issue because I somehow manually set the path to /config/transcodes in transcoding setting. I removed this path and now in the logs above jellyfin is writing to /cache/transcodes but the problem is still there.

[TheDreadPirate]

I see a bunch of a bunch of ffmpeg exited with code 137 after you switch to /cache/transcodes. Can you find one of those ffmpeg logs and share that? I've seen weird things happen when Jellyfin can't write to portions of its data or cache directories.

Can you restart Jellyfin and share your log after that completes? I want to see the Jellyfin startup logs.

[wzl778633]

OK so this is my process about how I got the logs below:
1. I restart the jellyfin server.
2. I opened a new website with chromium, login in to my account and directly choose the same movie to play.
3. I waited for the transcoding process until I can actually watch the movie "move" instead of the loading screen, and then I paused and exited.

Here is the logs from ffmpeg, I don't know why I can see 2 ffmpeg logs although I only played this movie once after the server rebooted.
https://privatebin.net/?24605d3d0738c528...eCRHPXbbcd This is the 1st one.
https://privatebin.net/?dd673f3bde10c8fd...ADBhjUj6vf This is the 2nd one.

Here is the logs from Jellyfin. Complete logs from startup.
https://privatebin.net/?b64e351c654d8997...j2xdXdhiYC