Configuring CMAN in RHEL 4

There are several variables you can configure in cman. These can be changed in cluster.conf, eg:

<cluster>
  <cman deadnode_timeout="100"/>
...
</cluster>

or numbers can be echoed into the appropriate proc file in /proc/cluster/config/dlm/.

The latter will take effect immediately (on that node only), whereas changing cluster.conf will only take effect the next time cman is started on that node.

The variables are:

joinwait_timeout

Time to wait while sending "Join Request" messages. This value should be at least twice the HELLO timer, probably 3 times. If a node does not receive a join_ack in this time it will give up trying to join a cluster. Join Request messages are sent to nodes in response to HELLO messages.

In seconds, the default is 16

joinconf_timeout

How long we wait after sending a "Join Confirm" message to a new node before regarding that node as died during joining.

In seconds, the default is 30.

hello_timer

The period between sending HELLO messages.

In seconds, the default is 5.

deadnode_timeout

If we don't get a message from a node in this period, then remove it from the cluster.

In seconds, the default is 21.

transition_timeout

The maximum length of time a state transition should take. If a transition takes longer than this then the nodes will elect a new master and restart the transition.

In seconds, the default is 15.

transition_restarts

Maximum number of transition restarts before we will leave the cluster.

The default is 10.

max_nodes

Maximum allowed number of nodes in a cluster.

The default is 128.

sm_debug_size

Size in bytes of Service Manager debug buffer

The default is 256.

newcluster_timeout

The amount of time to send NEWCLUSTER messages before assuming there are no other clusters (with our ID) on this network and forming our own.

In seconds, the default is 16.

max_retries

Number of times cman will resend a message before declaring it lost.

The default is 3.