Renaming Veritas Cluster Server Nodes

I recently found myself with the need to rename both nodes in my two node VCS deployment because they’d been built using the wrong naming convention. The OS on each node is RHEL 5.2 running VCS 5.0MP3.

The system hostname is unimportant to VCS. VCS uses the file /etc/VRTSvcs/conf/syname, which is created during the software installation, for use within the cluster configuration. To make the change at the OS level as well as the cluster level this is the procedure I followed. All commands were run as root. If any of the VCS related configuration files don’t have matching nodenames or the main.cf file becomes broken, you can be left with a completely broken cluster. Proceed with caution and check your changes before you save them.

Dump a known good copy of the cluster config and make the config read-only (From node 1)

/opt/VRTS/bin/haconf -dump -makero

Stop VCS on both nodes

/opt/VRTS/bin/hastop -all -force

Stop remaining Veritas Fencing Driver (Run on Both Nodes)

/etc/init.d/vxfen stop
/etc/init.d/llt stop
/etc/init.d/gab stop
/etc/init.d/vxpal.actionagent stop
/etc/init.d/vxpal.gridnode stop
/etc/init.d/vxpal.StorageAgent stop

After issuing the hastop command above, the stop on llt and gab aren’t required, but those modules didn’t unload after stopping the fencing driver and after I’d stopped them again. I began to stop other VRTS related processes and after stopping the three vxpal services all the drivers I was concerned about were gone. In this case I think the culprit may have been the gridnode driver. I may have been impatient, but this procedure won’t cause any harm.

Confirm that the vxfen, llt and gab kernel modules are no longer loaded

/sbin/lsmod | grep -e vxfen grep -e llt -e gab

Edit the following files on both nodes with the editor of choice. Change each instance of node1 and node2 with their respective new names

/etc/llttab
/etc/llthosts
/etc/VRTSvcs/conf/sysname
/etc/sysconfig/network
On node 1 ONLY
/etc/VRTSvcs/conf/config/main.cf

Verify config

hacf -verify /etc/VRTSvcs/conf/config

A reboot is required for the OS level re-naming

/sbin/init 6

After the reboot you’ll want to verify that VCS is running and that all resources configured in the cluster are operating as expected. For me a quick check is

/sbin/gabconfig -a
/sbin/lltstat
/opt/VRTS/bin/hastatus

Advertisement