Emby Server not releasing HW Transcoding streams

[iamspartacus 40]

Is this normal?  If I start a stream session which forces a transcode (HW) and then close the session (no active streams showing up in dashboard) I'm still seeing the transcode session processing in my GPU well after  (it's been 10+ minutes so far) the stream was closed.  Furthermore, I tried restarting the server and the transcode session picks right back up.  Again, no active streams.

 

 

What's the best way for me to produce clean logs for troubleshooting this?

Edited April 29, 2019 by iamspartacus

[Luke 42081]

Sounds like perhaps that data is just wrong. @@softworkz may have some insight.

[softworkz 5071]

Is this normal?  If I start a stream session which forces a transcode (HW) and then close the session (no active streams showing up in dashboard) I'm still seeing the transcode session processing in my GPU well after  (it's been 10+ minutes so far) the stream was closed.  Furthermore, I tried restarting the server and the transcode session picks right back up.  Again, no active streams.

 

 

What's the best way for me to produce clean logs for troubleshooting this?

 

What's your approach for 'still seeing the transcode session processing in my GPU' ?

 

Are you able to retrieve a list of running processes?

 

In this list, could you look for processes named 'ffmpeg' in situations where you think that transcode sessions would be progressing?

[iamspartacus 40]

What's your approach for 'still seeing the transcode session processing in my GPU' ?

 

Are you able to retrieve a list of running processes?

 

In this list, could you look for processes named 'ffmpeg' in situations where you think that transcode sessions would be progressing?

 

 

Do you see the attached screen shot?  It answers your questions.  That is a live look at my GPU processes that is in real time.

Edited April 29, 2019 by iamspartacus

[softworkz 5071]

No, it doesn't. Please answer mine.

 

PS: Please use a native OS-provided method to retrieve a process list.

[iamspartacus 40]

No, it doesn't. Please answer mine.

 

PS: Please use a native OS-provided method to retrieve a process list.

 

 

[softworkz 5071]

Thanks for the list. That's looking really bad...

 

What OS is this?

 

How did you install Emby?

 

Why is ffmpeg being run from '/bin/ffmpeg' ?

We're shipping a custom ffmpeg version and iirc, this is located in a different folder.

 

Would you be able to post a hardware detection log and just one of the ffmpeg logs?

 

Thank you very much.

[iamspartacus 40]

Thanks for the list. That's looking really bad...

 

What OS is this?

 

How did you install Emby?

 

Why is ffmpeg being run from '/bin/ffmpeg' ?

We're shipping a custom ffmpeg version and iirc, this is located in a different folder.

 

Would you be able to post a hardware detection log and just one of the ffmpeg logs?

 

Thank you very much.

 

 

I'm running the official emby docker container in Unraid.  I believe that's the location it runs ffmpeg from by default as I haven't customized anything.

 

Where can I find those logs?

[softworkz 5071]

No need to search through the file system.

 

Just go to the Emby Server Dashboard and choose 'Logs' in the left navigation panel (quite at the bottom)

[iamspartacus 40]

No need to search through the file system.

 

Just go to the Emby Server Dashboard and choose 'Logs' in the left navigation panel (quite at the bottom)

 

 

Is there any info I need to redact in these logs before posting?

[Luke 42081]

You can also send it via PM.

[iamspartacus 40]

No need to search through the file system.

 

Just go to the Emby Server Dashboard and choose 'Logs' in the left navigation panel (quite at the bottom)

 

 

PM'd.

[softworkz 5071]

Thank you very much for sending the log files.

 

What I can say is that everything seems to be fine about the ffmpeg location.  I'm not quite familiar with the Docker variant, so apologies for the confusion.

 

The ffmpeg log file appears to be interrupted all of a sudden. Do all those ffmpeg logs look like this?

 

Does hw transcoding actually work? Always? Or just with certain kinds of files? Or randomly? Or does it allways crash and fallback to software transcoding?

[iamspartacus 40]

Thank you very much for sending the log files.

 

What I can say is that everything seems to be fine about the ffmpeg location.  I'm not quite familiar with the Docker variant, so apologies for the confusion.

 

The ffmpeg log file appears to be interrupted all of a sudden. Do all those ffmpeg logs look like this?

 

Does hw transcoding actually work? Always? Or just with certain kinds of files? Or randomly? Or does it allways crash and fallback to software transcoding?

 

 

Most of the time it works fine.  This was the first time I noticed that it was holding onto the transcoding stream even after I stopped playing the file.  Though I can't be sure that it wasn't happening in the past because i'm not always monitoring my GPU usage.

[softworkz 5071]

I'm not sure if those zombie processes are really 'using' the GPU.

 

It rather seems that those ffmpeg processes have crashed for some reason but without actually exiting the process.

I come to this conclusion because all those remaining processes that we could see in the list that you posted are using the exact same amount of memory. That's no coincidence.

Instead, there must be a certain operation leading to a crash always in the exact same way. Otherwise they would show variable memory allocations.

 

In your screenshot, all visible instances were about playback of just a single movie.

 

It would be good if you could observe this situation for a while to see whether those problems could be isolated to a single video or a certain group of movies (e.g. HEVC or HEVC, 10bit).

 

But probably you should do a restart first, to get right of all that junk that has been aggregating so far..

[iamspartacus 40]

I'm not sure if those zombie processes are really 'using' the GPU.

 

It rather seems that those ffmpeg processes have crashed for some reason but without actually exiting the process.

I come to this conclusion because all those remaining processes that we could see in the list that you posted are using the exact same amount of memory. That's no coincidence.

Instead, there must be a certain operation leading to a crash always in the exact same way. Otherwise they would show variable memory allocations.

 

In your screenshot, all visible instances were about playback of just a single movie.

 

It would be good if you could observe this situation for a while to see whether those problems could be isolated to a single video or a certain group of movies (e.g. HEVC or HEVC, 10bit).

 

But probably you should do a restart first, to get right of all that junk that has been aggregating so far..

 

 

Will do.  I'm heavily testing Emby right now because I REALLY want it to replace Plex.  I'll have some more feedback in the coming days.

[softworkz 5071]

Great! Thanks for reporting.