does it means everything is ok

[Yuedong]

In my two node cluster,when the two nodes are on ,the serviceguard is ok and when shutdown one of the node,the pkg swicths well.

the only problem is that ,when the first node is
down ,and after i reboot the second node ,i cann't start the serviceguard in the second node (which means the cmruncl command cann't work in the second node),does it means there is something is wrong with my cluster,or means it's ok and i don't need to care about it.

[Duncan Edmonstone]

At initial cluster formation, all nodes must be present. So if one node is powered down and you reboot the other node it won't form a cluster automatically. This is a data integrity feature and the right way for the cluster to behave.

You can override this using the '-n' option on cmruncl:

cmruncl -n

But you need to be sure that the other node is down and not running the cluster.

See this for more details:

http://docs.hp.com/en/B3936-90122/ch07s02.html#startcluster

HTH

Duncan

[Yuedong]

thank you very much for your reply,i haven't run the command 'cmruncl -n hostname',but i have tried the command ' cmrunnode -n hostname' and it can't work. does it have any difference between 'cmruncl -n hostname' or 'cmrunnode -n hostname';

[Duncan Edmonstone]

Yes its different - cmrunnode just starts the cluster dameons on the specified node and those daemons wait for all other nodes to be present before joining the cluster.

cmruncl does the same on all nodes in the cluster during normal operation, but if you specify the -n it ensures that only the node(s) deatiled join the cluster.

HTH

Duncan

[melvyn burnard]

Yes, everything IS ok. SG i sbehaving as expected. At initial cluster start tim ewe require 100% of the nodes known about in the configuration to be available and ready to participate in cluster activities. If there is at least one node "missing" (as in perhaps powered off), then the other nodes will wait for a response for ten minets (600 seconds, the default cluster start timeout in the cluster ascii), then cease trying to form a cluster.
In this case (as in yours) the cmruncl command must be used to start the initial cluster with one or more nodes by using the cmruncl -n option, which overrides the requirement for 100% of the nodes.

The cmrunnode command is only used when you wish to have a node join an already running cluster

[Yuedong]

thank you very much, i'll test it in my cluster.