RE: dp264 disk or raid problem (Solved)

[Mark Hansel <hansel hansel mnstate edu>]

It was suggested I post a brief summary of the problem and the solution.
It turns out to be either a bad e2fsprogs rpm or some undetected
interaction between e2fsprogs and a RAID card not supported by RH. (Short 
message, use 1.27 release of e2fsprogs.)

Late last week, the machine crashed and rebooted with no log entries. It
rebooted with the (repair filesystem) prompt and enough information to
identify the offending partition as one on an ICP-VORTEX Raid5 logical
drive. (The ICP-VORTEX is not supported by RH, hence no aspersion intended
or appropriate.) fsck returend an hour of errors. Giving up on repairs, I
attempted to remake the file system (mke2fs). This failed with "short
write zeroing block) at the end of the file system (18 blocks from the
end). It failed on a shortened partittion (90%, 80%, 50% of original. All
failed at the same block from the end of the partition -- at least the
failure was consistent).

Why the crash and initial errors is a mystery that will remain unsolved.  
The frustrating part is the failure of both e2fsck and mke2fs to do the
right thing.

I did many tests (see earlier post for a partial list), including three
kernel releases, that led me to high confidence that the hardware was
good.

With help from some very generous community members, I verified that the
hardware was probably good and that the problem was software.
Specifically, e2fsprogs-1.26-1.71 did not work properly. 1.27, downloaded,
compiled and installed (over 1.26) worked.

I note that e2fsprogs 1.26 failed on an LX164 with no RAID and garden
variety SCSI hardware. A new cable/terminator was necessary to accomodate
an added drive, but I still needed 1.27 to make the file systems on the
new drive.

-- 
Mark Hansel
PO Box 41
Minnesota State University Moorhead
Moorhead, MN 56563
ph: 218-236-2039 fax: 218-236-2593
hansel@mnstate.edu
http://wwwcj.mnstate.edu