ZFS Pool Errors/high IO delay

[shactheorb]

a bit of a Proxmox noob so sorry if I'm missing anything, having trouble determining the issue here and appreciate the help.

I have a Raid 10 configured on my Proxmox server that's made up of 4 hard drive, 2 WD drives and 2 Seagate drives. A while back I was seeing some weird slowness issues with applications on the drives, and then saw that the ZFS pool was showing up as degraded. Specifically, one drive was showing up as Degraded, with the message "too many errors"


I had them hooked up to my server using a cheap NVME to SFF-8087 adapter, then a SFF-8087 to 4x SATA adapter, so assumed this was the issue. Replaced it with an nvme to sata adapter, and unfortunately that didn't change anything.

Here's a screenshot of the result from running zpool iostat -v as well, specifically SATA10 is the problem pool


Anything I can try to fix this? Hoping it's not an issue with the drive itself, but I can also look into buying a replacement if needed

 

[news]

You can replace "bad" drives, so by 4 new one:
Because
*) Seagate ST8000DM004 8 TB store data with SMR (Shingled Magnetic Recording), that's very bad!

Only use CMR Drives from Seagate NAS Pro

Change all your SATA III cable, buy new one and please check SATA III (600 mbit/s) with a clip rated.
Don't make sharp edge into the cables, use only a loose installation.

See zpool replace on openzfs wiki.

 

[UdoB]

I would like to add: whatever you do with rotating rust in the PVE context: add a fast "Special Device", using two mirrored SSD or NVMe.

 

[leesteken]

With SMR drives (or QLC SSD), I would expect write errors (due to timeouts caused by the bad performance). The checksum errors suggest that there is another issue like cables (as already mentioned), controllers or even memory (but unlikely with errors on one drive only).

 

[Kingneutron]

Seagate Barracuda drives are desktop-rated. You put them in a zpool with WD Reds, which look like they're CMR and probably OK.

Don't try running ZFS with desktop-rated drives, especially if they're SMR. If you like Seagate and want to stay with the brand, replace them one at a time with Ironwolf or Exos.

Basically, replace the Barracudas with NAS-rated drives and run a scrub; if the WD recovers then good. Otherwise you might RMA it.