SOLVED: Jellyfin in mixed HDD/SSD setup not working as expected

[Rezer]

Not sure if my expectations are flawed or not, but I have Jellyfin setup in docker with the cache, metadata, and config folders all setup to use a NVMe drive, while all the media is on the normal old fashioned spinning HDD array.  What I'm hoping to achieve with this configuration is to have it only access the NVMe drive when browsing the library, both for the speed benefit and to reduce the time the HDD drives are spinning.  The only time the HDD would be accessed is when media is selected and begins playing, allowing the drives to spin down for most of the day.   However, when browsing there's a fairly long delay when displaying each new page of thumbnails, and I can hear the hard drive chattering away.  Effectively this means any time anybody even glances at Jellyfin on their phone or whatever the hard drives have to jump into action, which as I understand things just shouldn't be the case.

The strange part is the cache, metadata, and config folders are all plainly visible on the NVMe drive, and no equivalent folders exist on the HDD, so what exactly would Jellyfin be reading from the media drive when simply browsing libraries?  And, more importantly, is there any way of stopping it from doing that?

[TheDreadPirate]

Are you using NFOs and saving images in the media directories?

[Rezer]

I turned off saving images in the media directories aleady, but NFOs are enabled in metadata savers in each library...I guess those are saved to and accessed from the media folder during browsing as well?

Edit: I turned off NFOs in each library, but the behavior is unchanged. Do I also need to delete the NFO files from each folder in the library? I wouldn't think so, but I have no idea at this point...

[TheDreadPirate]

If you save/have NFOs and images, it will read them from the hard drive (found this out the hard way).

So, yes, you will need to delete them. If your server is on Linux there are commands to automate this.

[Rezer]

Oh yeah deleting them isn't much of an issue, not a complete linux nub over here.  My thinking was just...if an NFO existing is enough to trigger a disk read, how can Jellyfin possibly know whether the NFO file is there or not without checking the disk beforehand anyway?  Kind of just dawned on me that the existence of an NFO can be stored in the cache, and it may then go try and retrieve the NFO file for any folder that it has marked as having an existing NFO.

Soooo, in that vein, I went into the library folders and deleted every single jpg and nfo file that was there, and to my surprise almost all the thumbnail images for the library disappeared.  Even for those that were added after I disabled "Save artwork into media folders"!  Currently rebuilding the image metadata, and thus far it doesn't look like it's outputting any jpg files to the library folders.  I'm thinking this was probably the issue, but won't be certain 'til it's done recreating all the chapter thumbnails.

[Rezer]

Yeah it seems to be behaving much better now after deleting all jpg, nfo, and png, and rescanning for missing metadata.  And hopefully it's not just because all the freshly regenerated images are still in memory... 

Thanks for the help!