[Linux-cluster] Node crashes when touching raw partitions

[Steve Nelson]

Hi All,

Would appreciate some advice...

I have a 2 node Oracle cluster - Red Hat Enterprise Linux AS release 3
(Taroon Update 6) // GFS 6.0

After an Oracle crash, the cluster has refused to come back up, and
investigations thus far show that when the secondary node attempts to
access the raw partitions (to obtain quorum information) the kernel
panics, and the node is fenced and rebooted.  This typically happens
when I try to start clumanager.

I have been unable to obtain any useful information to establish
exactly what is happening, and so am now trying to replicate the
problem with sg_utils.  Can anyone suggest methods to try to emulate
what happens when clumanager starts?

Any further suggestions would be most welcome.

Thanks,

S.