Description of problem: A couple of people on the list have seen this. cman seems to be quite happy ina cluster of 31 nodes, but adding the 32nd causes it all to fall apart. According to the one detailed log I have seen, the transition master node first falls out wih "Inconsistent cluster view". After that, the other nodes all leave with "No response to messages". I have not reproduced this yet (slight lack of real nodes) but I'm going to have a good try using virtualisation.
Just fot the record we (well kevin) have managed to get 39 Xen nodes running quite happily. So it appears there is no hard limit here at least.
We haven't been able to recreate this problem on the latest code base and the number of machines available at the time. We have run 39 virtual nodes in a cluster, so need help in recreating it. Marking the bugzilla as needinfo for now.
Marking this WONTFIX. Customers who need more than 31 nodes should use gulm or RHEL5.