On occasion, you may have cluster node hardware which have failed and won’t be replaced. As such, you will need to remove them as cluster nodes and remove any Logical Units that were being provided to them. This is accomplished in two stages.
1) Remove the cluster nodes (by forcing them out)
2) Remove the logical units
When the cluster has one or more cluster nodes down, running a cluster -status will indicate that the cluster is in “degraded mode”, and it will list the nodes in a “DOWN” state. If you attempt to remove the nodes, and they have any Logical Units (LUs) assigned, that will be indicated and you will be unable to remove the node. Thus, it will have to be forced.
Example command to remove a node from a SSP cluster.
cluster -rmnode -f -clustername cluster0 -hostname servera.example.com
To help track down any unused LUs, do a lu -list to get a list of all of the Logical Units. Then run lsmap -all on each of the running cluster nodes and make note of the backing devices. In my example here, there are two LUs which are unaccounted for, thus must of been assigned on the DOWN node.
To remove those LUs, we’ll run:
lu -remove -lu servera_rootvg
lu -remove -lu servera_appvg