Thanks for your quick response.
But we started the cman manually and then did not start any other service between.
There is no risk that the network goes down during the test.
We don't use actually 'intelligent switch' and no cisco switches.
We made a new test with the cman configured to start at boot time and with the clean_start set to 0.
1) launch cman on node 1
2) node 1 fences the node 2 as it was not joining the fence domain before the time out.
3) during the boot time of node 2, the cman is launched and then fences the node 1.
4) after the boot of the node 1, both seems to see each other with the correct state.
Is this is the expected behaviour?
do both nodes have to be rebooted before having all nodes ready?
Thanks a lot for your help.
Yes, I suspect the problem is that the node is 'bouncing' as it joins the cluster. Causes of this are usually to do with either a) startup scripts (eg some Xen ones) taking he interface down and then up after openais has started or b) "intelligent" switches taking too long to recognise the multicast join. So that both cluster nodes have "state" (the dirty flag) by the time they see each other. Chrissie