I have an issue which seems pretty weird to me, and I hope someone can shed some light on it.
We have a Glassfish cluster running Sun's Identity Manager application, and recently needed to migrate the cluster to new hardware. The version of the OS changed from Solaris 10 update 4 to update 6, so rather than try to export/import the zones directly, we removed the instance and nodeagent from the old zone followed immediately by the adding of a new node agent and instance on the new system. This worked well and after letting each instance "burn in" on the new hardware for a week we migrated the next one, and finally all 4 instances were migrated properly. The application works well on the new systems and there was zero downtime (yay, clustering!).
After a few weeks, we noticed a message in our error logs referring to the imq.cluster.masterbroker setting, which according to the startup logs for the MQ is different for each node in the cluster (each node shows the one "before" it in the list as the master: A->D, B->A, C->B, D->C). Since they don't match, the IMQ service doesn't start (we don't currently use it, but we are planning to in the near future).
We never set this setting manually anywhere (the MQ settings are all at the defaults: the first node should be the master broker according to the logs on our test cluster). How did this setting change, and how can we force it to be the same across the cluster? Is there a way to get it to go back to the defaults or must we do this manually in the IMQ config.properties file?
Thanks in advance,
Mark
[Message sent by forum member 'mcking' (mark_at_mccoyfam.net)]
http://forums.java.net/jive/thread.jspa?messageID=362014