sudo docker inspect sb-jellyfin-media -f '{{json .Mounts}}' | grep -i dri [{"Type":"bind",
Posted by Bigmoviehouse
My question: sudo docker inspect sb-jellyfin-media -f '{{json .Mounts}}' | grep -i dri [{"Type":"bind","Source":"/mnt/pool","Destination":"/data","Mode":"rw","RW":true,"Propagation":"rprivate"},{"Type":"bind","Source":"/opt/sparkbox/modules/media/config/jellyfin","Destination":"/config","Mode":"rw","RW":true,"Propagation":"rprivate"},{"Type":"volume","Name":"afe78a9b38731f504523540c560b2dc15c83c6c3eb38dfc73a25f9a24241ffb6","Source":"/var/lib/docker/volumes/afe78a9b38731f504523540c560b2dc15c83c6c3eb38dfc73a25f9a24241ffb6/data","Destination":"/cache","Driver":"local","Mode":"","RW":true,"Propagation":""}] chris@chris-OptiPlex-7060:/$ SparkBox AI replied: Confirmed. No /dev/dri in the mounts — the grep found nothing dri-related, just the same three mounts we've seen all along (/data, /config, /cache). So even after a full down/up, the hw overlay is not being merged into the Jellyfin service. That's the smoking gun, and it's exactly the bug we escalated. To answer you plainly: no, it's still not fixed. Whatever patch shipped didn't touch this. The state is: - /dev/dri exists on the host ✅ - JELLYFINHWACCEL=qsv is in .env ✅ - docker-compose.hw.yml exists ✅ - But after a clean recreate, Jellyfin gets no /dev/dri device, no PUID/PGID/env from the overlay ❌ This is the overlay merge failing, and it's not something you can fix with set-env, up, or any manual edit (manual edits get stomped on the next update anyway, and you'd be fight …(reply truncated — full chat sent to Tom) docker-compose.hw-nvidia.yml Aug 2 (hyphen) docker-compose.hw.nvidia.yml Jul 4 (dot) docker-compose.hw.yml Jul 4
1 replies
Chris wrote:
Thanks for laying this out so clearly — it saves a lot of digging. You're right, and it's on our side, not your setup: Jellyfin isn't being handed your Intel GPU the way it should be, even though the GPU is there and the setting is on. That's a known trouble spot we're actively working, and the dashboard AI has already sent the full details over to Tom. I've logged this thread too so it's tracked. There's no command on your end that forces it through right now, so please don't burn more time on it — we'll post back here when there's a real fix.