Cluster Quorum Disk

[simon_164]

I want to shutdown one node of a two node cluster , i'm using also a quorum disk. how can i be sure that this is not the node that is responsible of the cluster. I mean how can i be sure that if i shutdown the node no disk should be unmounted/mounted. i want to know what is the node that is active within the cluster.

[Uwe Zessin]

All cluster members are "active" (= permanently talking to each other) in a VMS cluster. If I remember correctly, disks are only dismounted if somebody has put commands into SYSHUTDWN.COM and/or the REMOVE_NODE option has been used.

If this is a two-node cluster with a proper setup of VOTES and a fully shared-disk subsystem, you can safely shutdown one node and the other will continue.

[simon_164]

isn't the disks in an openvms cluster locked by one node and whenever this node is shutdown, the disks should be unmounted and then mounted on the other node. Isn't this the case?
can both nodes access the same disks at the same time? what about data corruption?

[Karl Rohwedder · Accepted solution]

In an OpenVMS cluster all disks may be mounted on all members at the same time, accessing them at the SAME time. This is not Windows...

If not all nodes have direct access to the disks (this is called serving them from one node to other nodes), then, if you shutdown the serving node, the disk will go into a state called mount-verification on the other nodes. All IO is halted until the serving node is rejoining the cluster. Then everything will continue.

regards Kalle

[Uwe Zessin]

OpenVMS has a shared file system since the introduction of clusters about 20 (twenty) years ago! It allows concurrent access of volumes right down to the record level.

[simon_164]

So i guess my disks shouldn't be served. and the quorum disk is shared and accessed by all the nodes.

[comarow]

The question is, where is the qurom disk?
It SHOULD be on a shared controller, not directly connected to any system, (like a direct connected SCSI disk),

If it is on a SAN or SCSI controller, like an HSZ, or an older HSJ you have no problem.

If you do a show dev/full on the quorom disk, we can know for sure.

I'm assuming your voting scheme is
each node has one vote and the quroum disk has one vote. This assures you can take down either node in the cluster. I assume you are indeed using the quorum disk. To verify this type

$show cluster/continuous

Then, while that is displaying type

add qf_vote

a box will appear and it will say either
YES or NO

If it says YES you are using your quorum disk.

If it doesn't, assuming you have support, log a call with 800-633-3600 and ask for the VMS team so they can walk you through it.

Best of luck

Bob Comarow