FreeBSD ZFS: Performance degration, disk replace
Christian Kruse,
My home NAS is based on a FreeBSD server with root ZFS on 4 physical disks and two disks redundancy (raidz-2). I'm pretty satisfied with this solution until a few weeks ago where the performance of the raid degraded. It was really unbearable. After googling a bit I found several tuning guides, but none of them helped. But one of them suggested to check the drive health, and smartctl -a showed that one of my disks had read errors. To test this I took the disk out of the array:
$ zpool offline zroot ada2p3
NAME STATE READ WRITE CKSUM
zroot DEGRADED 0 0 0
raidz2-0 DEGRADED 0 0 0
ada0p3 ONLINE 0 0 0
2021608558689218476 OFFLINE 0 0 0 was /dev/ada2p3
ada3p3 ONLINE 0 0 0
Immediately the read and write performance increased. So I bought a new disk, copied the partition table via
$ gpart backup ada0 | gpart restore -F ada2
$ gpart bootcode -b /boot/pmbr -p /boot/gptzfsboot -i 1 ada2
And re-attached it with the following command:
$ zpool replace zroot /dev/ada2p3
Now my raid is rebuilding:
pool: zroot
state: DEGRADED
status: One or more devices is currently being resilvered. The pool will
continue to function, possibly in a degraded state.
action: Wait for the resilver to complete.
scan: resilver in progress since Wed Dec 5 10:25:12 2012
35.3G scanned out of 2.00T at 67.4M/s, 8h28m to go
8.79G resilvered, 1.73% done
config:
NAME STATE READ WRITE CKSUM
zroot DEGRADED 0 0 0
raidz2-0 DEGRADED 0 0 0
ada0p3 ONLINE 0 0 0
ada1p3 ONLINE 0 0 0
replacing-2 OFFLINE 0 0 0
2021608558689218476 OFFLINE 0 0 0 was /dev/ada2p3/old
ada2p3 ONLINE 0 0 0 (resilvering)
ada3p3 ONLINE 0 0 0
Yay, that was easy! I really like ZFS.