I get periodic I/O errors on a couple of SCSI partitions that are part
of RAID1 devices, eg:
SCSI disk error : host 0 channel 0 id 0 lun 0 return code = 28000002
extra data not valid Current error sd08:08: sns = 70 4 ASC=44 ASCQ= 0
scsidisk I/O error: dev 08:08, sector 5603386
...
Usually the failure looks like this in /proc/mdstat:
md3 : active raid1 sdb8[1] 6184896 blocks [2/1] [_U]
ie: sda8, the other mirror, is not listed at all.
I just run raidhotadd /dev/md3 /dev/sda8 to recover.
But this time, it went like this:
md3 : active raid1 sdb8[1] sda8[0](F) 6184896 blocks [2/1] [_U]
and raidhotadd said:
/dev/md3: can not hot-add disk: disk busy!
both before and after umount.
So, umm.., raidstop, raidstart, raidhotadd (recover, rec...), mount,
and ASW for another day or 3 :)
Can anyone explain these different failure modes?
--
2.2.10 (SMP) with the 7/24 raid0145 patch and v 0.90
of the raidtools w/autodetect.
Supermicro P6DBU w/onboard AIC7xxx. dual P3.
drives are IBM 9GB DRVS09D LVD 10k.
(Yeah, I am gonna upgrade to ac's 2.2.12-final RSN).
--
TIA, Will Brown
