Shutting down your
Open VMS Cluster
Nicely

Sometimes you want to bring your Open VMS Cluster down in a neat and orderly fashion. If you have a large Cluster with multiple boot nodes, and multiple CPU architectures, this may become difficult.

You may want to bring down certain groups or classes of nodes one at a time, finally leaving your boot nodes, and then bringing them down.

The VMS SYSMAN utility has the feature of shutting down the whole cluster, but this can cause problems if the boot server has some serving process that shuts down before the satellites can properly close their connections. The SYSMAN utility can also use logical names to designate classes of nodes for shutdown or reboot, but this too can cause problems if one or more nodes in a designated class is already down.

That is why I have written the command procedure CLUSHUT.COM. This utility command procedure looks at the currently running cluster, and builds classes of node lists dynamically. You can then specify your own list of nodes, or select one of the predetermined node classes, then have those nodes shutdown or rebooted.

Once your list or class of nodes has been specified, each node is tested to see if it is a valid member of the current cluster, and that it is up and available to be shut down or rebooted. If all is well, you are presented with a list of the nodes, and once again asked if you want to continue with the actual operation.

All shutdown/reboot options that are asked by the default SYS$SYSTEM:SYSHUTDOWN.COM command procedure are also asked in the same manner as the normal shutdown procedure, with the exception of the POWEROFF option. This is because of the lack of the POWEROFF option being implemented in the SYSMAN utility as of VMS version 7.2. If, and when, the POWEROFF option is implemented in SYSMAN, then just un-commenting two flagged lines in the command procedure will allow the implementation of the POWEROFF option (assuming the SYSMAN standard is followed).

After you download CLUSHUT.COM, place it into the common root of your system manager directory ([VMS$COMMON.SYSMGR]) for each boot server system disk in your cluster.

Because there is no good, completely transportable method in Open VMS to determine if a particular node in a Cluster is a Boot Serving node,you have to do this manually by tailoring the first executable DCL line after the comment section in the command procedure CLUSHUT.COM.

The first executable line after the comment section in the file says :

$     BOOT_NODES = "KIRK,SPOCK,MCCOY"

You must change it to list all of your boot serving nodes in your particular cluster.

Once you have made the change, and placed the file in the proper directory on all of your system disks, then you are ready to go.

Of course, if you have any questions about this or any other of my utilities or instructions, pleas feel free to eMail me.