Hi there!
we recently replaced all 6 HDs (80GB) of our ide-scsi hardware
raid5(ift-6200) by 160Gb ones.
This Raid contained 2 Partitions (each ~200GB). I backed up those partions
using tar to another ide-sci hw raid5(ift-6300, 1.7TB,12x160GB) without
errors.
Now, while migrating the data back, I get file-system corruptions on the
newly created xfs-filesystems.
tar xvfp scratch.tar stops with input/output errors.
I'm using cvs-kernel checked out on August, 4th, ca. 18:00 EST, compiled
with gcc-2.95.4(DebianPrerelease).
xfsprogs are the latest, too.
According to the manufactor of the ift-6200, I should use 32k stripe unit
for random-access optimization. So here's what I used to create the two
filesystems on the ift-6200:
mkfs.xfs -f -d sunit=64,swidth=320 -l size=32768b /dev/sdb1
mkfs.xfs -f -d sunit=64,swidth=320 -l size=32768b /dev/sdb2
Should I try other options? e.g no sunit and swidth??
The attachment contains the whole dmesg output.
thanks in advance
Christian Guggenberger
dmesg.out
Description: Binary data
|