Strange cluster behaviour

[Bruno Bossier_1]

I have a 2-node cluster, let's say node A and B and 1 fail-over package.

Package runs on node A.

When I execute a shutdown -h 0 on node A, package switches over to node B. This is OK. Then I shutdown -h 0 node B. All still OK. Now I restart node A. Cluster should normally come up, but it doesn't. From syslog.log I can see there was an attempt to start the cluster (as part of /sbin/init.d/cmcluster) but it fails.

Any ideas ?

[Pedro Cirne]

Hi,

If one of the nodes is down, you have to starts the cluster manually by executing "cmruncl" on the node you boot.

To start automatically, both nodes must boot.

Enjoy :-)

[Duncan Edmonstone]

Bruno,

This is normal behaviour - all cluster nodes must be present when a cluster starts - its to prevent split brain syndrome.

Obviously when a cluster reforms rather than starts this is not the case.

If you are absolutley *sure* that the other node is in fact down, then you can start the cluster on a single node using 'cmruncl -n nodeA'

HTH

Duncan

[Bruno Bossier_1]

OK, thanks to all for the clarification !

Cheers,
Bruno