HASingleton quorum?

[mpatercz]

Hello there!

I'm considering a following scenario: there is a connection issue between nodes in a cluster, causing the cluster to split. Does that mean my singleton service will be deployed on more than one node (since we have more than one cluster now)? If so, how can I prevent this? Is there a quorum setting, which would allow only a majority of nodes to elect a master? It is accepted for the singleton service not to be deployed at all when the cluster splits, the more important thing is to prevent it from being deployed more than once.

Thanks in advance,

Marek

[pferraro]

By default, a cluster split will indeed cause a new singleton master to be elected within each partition, until they are mended.  To override this behavior, you can supply a custom election policy (i.e. implements org.jboss.ha.framework.interfaces.HASingletonElectionPolicy) that requires a quorum. 

[mpatercz]

Hi Paul, thanks for the anwser. That looks like the way to go.