SOLVED: Integrated GPU Irregularly Invoked, Playback Issues Result

[Brian]

BLUF: 

Beginning with v10.9.7 and continuing through current v10.9.10, my Jellyfin server no longer reliably invokes integrated GPU resources (Intel UHD Graphics 770) at every playback request. If not invoked, during playback of recordings the progress and time remaining timestamps are nonsensical and fast-forward and rewind functions do not work. This post reflects additional observations and conclusions concerning playback issues about which I have posted previously.

Details:

Whether the GPU is invoked can be determined at recording launch using sudo intel_gpu_top. If properly invoked, the GPU generates numerous 3MB .ts files in the /var/lib/jellyfin/transcodes folder. If these transcodes files are generated, the time markings displayed during pause, fast forward, and rewind are accurate - the bottom left corner shows progress through the recording, while on the right time remaining, in the format hr:min:sec. If time is accurate, fast-forward and rewind behave as expected.

If the GPU is not invoked, the small .ts files are not generated, and I assume playback is driven directly off the single very large .ts file generated at recording; The timestamps are nonsensical extremely long numbers, or if segmented into hr:min:sec, a one hour recording appears to be on the order of 4-5 hours.

Why the GPU is no longer invoked in all instances is a mystery. All source recording files are .ts. I've even seen instances when a particular recording is sometime replayed properly, and other times not. I've monitored the logs in /var/log/jellyfin but have not recognized any indication why the GPU is not being invoked.

Is anyone else experiencing this, have a suggestion for changing a setting or otherwise resolving this, or should I file this as a bug report?

Thank you.