I have a Solaris 10u3 x86 patched up with the important kernel/zfs/fs
patches (now running kernel 120012-14).
after executing a ''zpool scrub'' on one of my pools, i see I/O
read errors:
# zpool status | grep ONLINE | grep -v ''0 0 0''
state: ONLINE
c2t1d0 ONLINE 9 0 0
c2t4d0 ONLINE 32 0 0
c2t5d0 ONLINE 7 0 0
Are these errors important enough to switch the disks?
if not, i''ve read that when these numbers break a magic threshold, zfs
takes the disk offline and calls it dead.
If I use ''zpool clear'', will only these administrative
statistics be
cleared, or will important internal numbers that keep track of the
errors be cleared as well?
I do see bad blocks on the offending disks -- but why would zfs see them
(assuming the disk re-mapped the bad blocks) ?
# smartctl -a /dev/rdsk/c2t1d0 | grep defect
Elements in grown defect list: 3
# smartctl -a /dev/rdsk/c2t4d0 | grep defect
Elements in grown defect list: 3
# smartctl -a /dev/rdsk/c2t5d0 | grep defect
Elements in grown defect list: 2
--
Jeremy Kister
http://jeremy.kister.net./