[Linux-cluster] SCSI fencing on RHEL 4.5

[Maciej Bogucki]

Sadek, Abdel napisał(a):
> I have a 2-node RHEL 4.5 cluster. I am using SCSI fencing which uses
> scsi-3 reservation. The 2 nodes are sharing devices on a FC storage
> array. I am using a single path; there is no multipath driver involved.
> I am also using GFS on top of LVM volumes.
> 
> When I shutdown one of the nodes, the cluster does not send any commands
> to the storage array to clear out the persistent reservations for that
> node. This causes the resources not to fail correctly to the other node
> since it cannot break the reservation done by the dead node. My
> questions are:
> 
> 1. Is that a known issue?
> 
> 2. Is there a way to turn some kind of debugging on the cluster that
> will allow me to watch if any scsi PRIN/PROUT commands are sent to the
> storage devices?

Hello,

1. Check in Your logs if cman run /sbin/fence_scsi
2. If Yes, check why this scritp failed fe. white wrapper script
3. If no, check why cman doesn't do fencing

Best Regards
Maciej Bogucki