VA connectivity failed from node1 which contains Cluster lock failed

[HP-UX admin]

Hi,
Recently I faced one problem with serviceguard .My Cluster lock disk is in VA7410 & from node1 we suddenly lost communication with VA.So in syslog it was giving message that Cluster lock disk failed it u any node faile cluster will crash.
Since VA not accesible from node 1 cluster & package was not able to halt.Also I cannot able to shutdown the nodes because since cluster lock not available the cluster could crash.
What is required to do in this situation.pl guide.

[Stephen Doud · Accepted solution]

The packages could not halt because they apparently had file systems that were mounted to the VA7410. HP-UX cannot umount a file system that is not reachable, hence, the package halt cannot complete successfully. In turn, the package control script cannot deactivate such VGs, so the Serviceguard cmlvmd cannot sign off because it still registers those VGs as open.
Therefore, in order to get the node to leave the cluster, a reboot is necessary.

However, if the cluster lock VG is not accessible to the other node, it will not be able to access it in the cluster reformation process - and will be forced to TOC.

My recommendation is to repair the cluster lock disk - which may require a shutdown of both nodes, if the disk is not hot-swappable.

If the repair requires a replacement disk mech, contact the HP Support Center to acquire the 'cminitlock' utility that will restore the cluster lock structure while the cluster is running. This will allow lock disk arbitration when needed.

[HP-UX admin]

Thanks STEPHEN.

I cannot replace cluster aware disk because this device is LUN created space in VA7410.
Since VA communication failure frm NODE1 Cluster lock is not available frm node1 but it is available frm node2.
HOW to shutdown nodes? One at a time or simultaneous because after shutdown it is possiblity to crash Cluster.

[Stephen Doud]

cmhaltnode informs the other servers in the cluster that the node is leaving the cluster... therefore the remaining nodes do not need to access the lock disk to arbitrate the new cluster membership. However since node1 cannot perform a cmhaltnode because it cannot access the lock VG, I suggest you halt node2 first, then run 'shutdown -ry 0' on node1.

[HP-UX admin]

HI,
when cmhaltnode is done on node2 it will try to switch these packages to node1 which is not available in cluster so is it create any problem?