Hello zfs-discuss,
S10U5+patches, SPARC, Sun/qlogic 4Gb dual ported fc cards.
ZFS does mirroring between two lun''s, each is a lun comming from
separate 6540 disk array.
I got a kernel panic while pool was imported on one of the nodes
(kernel panic - it''s my fault). After reboot pool was imported
however it was marked as degraded and one lun was marked as
unavailable. If I run format on a unavailable disk I could read
the label but it was definitely garbled. On the other node
slice layout for both luns (standard EFI layout for zfs) were
ok. When I tried again I got warning that I need to use fdisk on
a disk... So i exported the pool and imported on the other node. Pool
imported fine with both devices, no errors. I exported it again
and tried to import on first node. Same story. So I exported it,
uncofnigured devices via cfgadm, did devfsadm -vC, checked with
format - and this time it could see probel labels/slices on both
drives, imported pool without any issues. Now I run scrub and it
detected over 5k CKSUM errors on that previously unavailable
disk. I run scrub couple more times and no more errors.
What bothers me is why did I got CKSUM errors? Looks like
something went terribly wrong outside of zfs (due to label
issue).
Unfortunately I didn''t have enough time to investigate it in
more detail, I needed to quickly get it fixed.
--
Best regards,
Robert Milkowski mailto:milek at task.gda.pl
http://milek.blogspot.com