I’m adding in an older Hitachi 1TB desktop drive to my FreeBSD ZFS server. I want to get it paired up with another Hitachi drive that’s in use on the system.
Since I want both of the Hitachi drives to be my mirrored drives for my “backup” pool and filesystem (where our home laptops drop their backups), and one of them is a mirror in another pool, here’s what I’m doing:
- Add the Hitachi drive as an additional mirror (creating a 3-way mirror) on the backup pool
- After the mirror above is synced, remove a Samsung drive from the backup pool
- Add the Samsung (removed from the backup above) to the data pool mirror that has the other Hitachi drive
- After the data pool is sync’d, remove the existing Hitachi from the data pool and put it as a mirror in the backup pool
- After the backup pool is mirrored up, remove the remaining Samsung drive from the backup pool and assign it as a global spare, leaving two matching Hitachi drives in the backup pool
During step one, I noticed that one of my two Samsung drives in the ZFS mirror is reading at double the speed of the other.
These drives are the same model, firmware, everything. What would make ZFS try to read twice the data off one drive vs. the other? Could one drive be slower than the other do to some problem with the drive, and ZFS is adapting by reading more from the more responsive drive?