[Lustre-discuss] reboot of OSS and OSTs are missing

[stumped]

I'm having a tough problem where a reboot of an OSS results in error
like:


e2label: Bad magic number in super-block while trying to open /dev/
mapper/ost_oss01_lustre0102
Couldn't find valid filesystem superblock.

The systems have HBAs, connecting (at the moment at least) to a single
nexsan sata beast (which has 6 6TB volumes).

I have multipath configured to create the above alias for the more
generic mpathN.  It did this before setting up aliases.

I'm running SLES 10.1.

Before the reboot, blkid found the paths to the device as:
/dev/sda:  LABEL="i3_lfs1-OST0003" UUID="555ce367-a02d-4826-
a0e7-91633e9687df" SEC_TYPE="ext2" TYPE="ext3"
/dev/sdd:  LABEL="i3_lfs1-OST0003" UUID="555ce367-a02d-4826-
a0e7-91633e9687df" SEC_TYPE="ext2" TYPE="ext3"
/dev/dm-0: LABEL="i3_lfs1-OST0003" UUID="555ce367-a02d-4826-
a0e7-91633e9687df" SEC_TYPE="ext2" TYPE="ext3"

Strangely, the filesystem which is used by the MDS/MGS is immune from
the issue.  It's also running SLES with a qlogic HBA.

I'm using lustre version 1.6.4.2.

Any ideas?

Thanks much,
JR