Removing a node from an active VCS cluster

In my previous article we discussed the steps need to add a new node into an existing VCS cluster (read more). In this article we perform the opposite by removing a node from the cluster.

This task can be performed only without interrupting the customer services. However you should make sure that you have sufficient resources to handle the load of the node being removed from the cluster.

For this example we will be removing vcsdevnod03 from the cluster

Stop VCS on the node

Top stop VCS on vcsdevnod03 we need to evacuate it, so that any service group running will switch over to the other nodes:

# haconfig -makerw
# hasys -freeze -persistent vcsdevnod03
# haconf -dump
# hastop -sys vcsdevnod03 -evacuate

Stop I/O fencing

Stop the VCS I/O fening on vcsdevnod03

# /sbin/vxfen-shutdown
# vxfemadm -d

NOTE: Check status of fencing from vcsdevnod01 or vcsdevnod02

Stop LLT/GAB modules

Stop the LLT/GAB modules on vcsdevnod03:

# gabconfig -U
# lltconfig -U

Remove VCS software

Remove the VCS software from vcsdevnod03 by running the installer utility and rename the file

Modify AutoStartList and SystemList

Run the following commands on vcsdevnod01 and vcsdevnod02 (assuming both nodes have vcsdevnod03 listed in their SystemList and AutoStartList attributes)

On vcsdevnod01

# hagrp -modify SG01 AutoStartList —delete vcsdevnod03
# hagrp -modify SG01 SystemList —delete vcsdevnod03

On vcsdevnod02

# hagrp -modify SG02 AutoStartList —delete vcsdevnod03
# hagrp -modify SG02 SystemList —delete vcsdevnod03

Remove node from ClusterServices SG

Remove vcsdevnod03 from ClusterServices SG on both vvsdevnod01 and vcsdevnod02

# hagrp -modify ClusterService AutoStartList -delete vcsdevnod03

Remove node from cluster

Now remove vcsdevnod03 from the cluster and sve the configuration (run this on vcsdevnod01 or vcsdevnod02)

# hasys -delete vcsdevnod03
# haconf -dump -makero