upgrade to 6.12.3 + ZFS cache for docker = sluggish apps

[ShadowLeague]

Greetings,

 

I made the jump from 6.11.5 to 6.12.3. After the upgrade, I converted my two cache pools to zfs. The cache pool running my docker containers use two NVMe drives in a raid0 (yep - I know the risk and make backups  ) Since converting this to a zfs cache pool, my containers seem to be a bit sluggish. For example, Plex spins when I try to Navigate from my TV library to recommended TV items, or just times out altogether. I've observed WebUI timeouts with SABnzbd. Troubleshooting steps:

• restart unRAID
• stop/start docker (ver 20.10.24)
• in Docker, switch from macvlan to ipvlan
• FYI, Docker info still shows the filesystem as btrfs
• in Plex, refreshed metadata
• ensured appdata folder sits on the zfs cache pool, and no secondary storage is defined. 

 

While my system isn't new, it isn't starving for resources (I think). I have two E5-2600's v3 with 64GB of ECC Ram. If I look at the unRAID dashboard for RAM, ZFS is taking about 8GB, and Docker is steady at 41% utilization. 

 

Is there anything else I can try? Is there a step I have missed?

Edited August 16, 2023 by ShadowLeague
added additional information on Plex experience

[ShadowLeague]

Might have made progress:

 

• Deleting docker apps and re-adding them didn't help
• Deleted the docker.img file and downloaded a fresh copy
  • reinstalled docker apps via community apps plugin (thank to the author(s) for this)
  • so far, the lag experienced in Plex is gone, but more testing to follow

Maybe the old btrfs settings were causing an issue? I also just specified an image folder and not a docker image... not sure if that made a difference.

[Teffi71]

@ShadowLeague could you provide a screen shot of your shares?
I have also two NVMe's which I want to setup up for docker and the other for VM's.

TIA

[ShadowLeague]

For Plex, it's back to taking awhile to view recommended items. It might take 10-15 secs before anything is displayed. After that, it works as expected. Other times, it'll time out. I'm optimizing the database to see if that helps. I'm not seeing a delay in SAB but I don't feel like I've tested it enough to confirm it's fixed.

[ShadowLeague]

14 hours ago, Teffi71 said:

@ShadowLeague could you provide a screen shot of your shares?
I have also two NVMe's which I want to setup up for docker and the other for VM's.

TIA

Are you wanting help setting them up or you're experiencing the same issue?

[ShadowLeague]

optimizing libraries had no effect. I'm going to rebuild the cache array to something other than zfs to see if that helps.

[ShadowLeague]

I converted my cache pool to btrfs with Raid0 and Plex and other app UIs running from this pool are back to being responsive. 
Someone on reddit mentioned specifying the docker image (and not the folder location) helped with their issue. At this point, I'm happy with the performance and I don't think it's worth further testing. I did keep my second cache pool as zfs and don't have any issues with it.