Testing Cluster Halt and Reconfiguration

This section shows how to test the correct reconfiguration of the cluster after the loss of a node. To do this in SAM, read the next section "“Using SAM to Test Cluster Halt and Reconfiguration ”." If you want to test reconfiguration using HP-UX commands, skip ahead to the section entitled "“Using HP-UX Commands to Test Cluster Halt and Reconfiguration ”."

Using SAM to Test Cluster Halt and Reconfiguration

Perform the following steps:

From SAM, select the High Availability options, then choose Cluster Administration.
In the Cluster Administration area, choose Specify Node(s) to Leave the Cluster, and then select one node to halt. Reply Yes when asked for a verification.
When the Cluster Administration area screen reappears, make sure the selected node is no longer an active part of the cluster.
From the Action list, choose View Syslog. Read the messages to verify that the reconfiguration has taken place.
From the Action list, select Specify Node(s) to Leave the Cluster, and choose the other node. Reply Yes when asked for a verification.
In the Process Management area of SAM, observe that none of the following daemon processes are running:
- cmcld - CM daemon
- cmgmsd - group membership daemon (OPS 8.1.6 and later)
- cmlvmd - SLVM daemon
These are shown as children of the init process.

To start the cluster running again, in the Cluster Administration area, choose Start Cluster, and select All Nodes.

Using HP-UX Commands to Test Cluster Halt and Reconfiguration

To test the correct reconfiguration of the cluster following the loss of a node, issue the following sequence of commands on node 1:

# vgchange -a n /dev/vg_ops 
# cmhaltnode -v

This stops node 1. Use the following command on node 2 after waiting about a minute for the reconfiguration to take place:

# cmviewcl -v

The output of the command should show that the cluster has reconfigured with only a single node.

NOTE: If you issue the cmviewcl -v command on node 1 after halting node 1, you will see the following message:

CLUSTER      STATUS
cluster1     down

If the cluster is running, always be sure to issue the cmviewcl command on a node that is an active participant in the running cluster.

Use the following command on node 2 to halt the second node:

# vgchange -a n /dev/vg_ops 
# cmhaltnode -v

The use of cmviewcl -v on either node should now show that no cluster is active. At this point, you can use the ps -ef command on both nodes to show that the following processes no longer exist:

cmcld - CM daemon
cmgmsd - group membership daemon (OPS 8.1.6)
cmlvmd - SLVM daemon

To start the cluster is running again. Use the following commands on each node:

# cmrunnode -v 
# vgchange -a s -p /dev/vg_ops

Testing Cluster Halt and Reconfiguration

Technical documentation

» Table of Contents

» Index

Using SAM to Test Cluster Halt and Reconfiguration

Using HP-UX Commands to Test Cluster Halt and Reconfiguration