Tdarr: TomAI Flag
Posted by kiltmann
TomAI: I'd flag this one — you've got clean, reproducible proof (.env set correctly, hardware confirmed via lspci, mount still missing after restart) that Tdarr doesn't get the same hw passthrough Jellyfin does. So the gap is real: the hw overlay logic appears to only target Jellyfin's container, not Tdarr's — even though Tdarr's own guide text talks about setting JELLYFINHWACCEL for hardware transcode plugins. That's inconsistent, and it's not something a restart or config tweak on your end fixes — this needs an actual overlay fix from Tom shipped in an update. Thanks! Klint
4 replies
Chris wrote:
Thanks for flagging this — good digging. One correction on the Tom AI answer though: the code actually does wire up hardware transcoding for Tdarr the exact same way as Jellyfin (same JELLYFINHWACCEL setting, same /dev/dri passthrough), so it's not a case of Tdarr being left out. That means what you're hitting is more likely a real glitch than a missing feature. Two quick checks: make sure you restarted with sudo sparkbox restart tdarr specifically (not just media), then run sudo docker exec sb-tdarr ls /dev/dri right after — if that comes back empty, paste it here and I'll get it in front of Tom.
kiltmann wrote:
sudo docker exec sb-tdarr ls /dev/dri card1 renderD128 I had previously changed JELLYFINHWACCEL to qvs (qsv?) and just returned it to auto. The did sudo sparkbox restart media. The above is the result.
Chris wrote:
Good news — that fixed it. card1 and renderD128 showing up means the GPU device is there now, so the earlier missing-after-restart issue is resolved, and auto works fine for detecting it (no need to force qsv by hand). One more step though: having the device present doesn't automatically make Tdarr use it. Open your transcode plan in Tdarr itself and make sure a hardware (QSV/VAAPI) plugin is selected there too, not just the software one — that's the actual on/off switch for GPU transcoding.