Interesting. I will say music is the weakest point of jellyfin db. Mine is currently good but I nuked my db for the new db style in 10.11 and created a new db, mainly bc of issues within music library. But this went back to like 10.9 and were probably from me fucking around in the db manually to try and override tags. I’ve since retagged my music as it was just the easier solution (artist ft artist got me all fucked up, I hated having 900 entries for like 1 artist who did frequent collabs)
https://github.com/jellyfin/jellyfin/issues/15070
This is the issue thread re the very long scanning after upgrading to 10.11
Edit: also one thing I learned from the db editing and again from the 10.11 migration is that the jellyfin db can get “stuck” in such a way that nothing I can find will overwrite db entries, even manually opening the db in an sql editor (which makes me think there are hard to find tables I could never track down but db stuff is not my forte so I dunno)
Oh right! In admin console under server>general at the bottom there is a section called “performance”. Under this the first entry is “parallel library scan tasks limit”. Increasing this can significantly improve performance here. If it’s empty it is auto selecting. To that point if you’re running jellyfin in a docker and only allow it access to one core that would seriously limit performance here. Screenshot below, I currently have it set to 2 because I’ve done the “big” scan and I don’t need my cpu hammered when jellyfin does library scans
Additionally an issue I ran into later: under playback>transcoding enable hardware transcoding must be checked. Trickplay task was taking literal months because even though I had configured igpu transcoding correctly this has to be enabled separately. UX stuff like this is where jellyfin needs more polish; I’m sure there’s debate about this but why would this not default to on once transcoding is enabled? Who knows. Maybe just oversight.