I have some 8TB WD red 5400RPM drives in an HP N54L running OpenIndiana.
When two drives are in a mirror configuration and passing a constant stream of data (transferring 6TB to the newly created pool), ZFS will fail them with write errors. When in a RaidZ, it seems to be happy, as if the data transfer is slowed down enough for the drives to cope. As well as a 6TB sustained transfer, three scrubs in a row in RaidZ went through and found no errors.
I’ve gone through a small number of drives with different suppliers, and I’m now convinced that this is a drive problem, and RMAing and swapping drives will not solve this issue.
Indeed, when I commented what I was going through on my social media, I received comments like…
“I had terrible luck with WD Red 8TB with ZFS. Are they Shingled Magnetic (SMR) Drives? I had to increase several write timeout parameters to stop getting errors with them.”
Obviously, there are no tools on OI to address drive parameters, and being 8TB they were outside the shingle issue, so I’m not sure what to do about this.