Every once in a while, it will be noted that there is an OBJECT_NOT_EXIST BAD_SERVER_ID in the nodeagent traces which could make the underlying EJBs on that node unavailable. Now this may affect just one JVM, on the node. YOU could have multiple JVMs on a single node.
Symptom : Client gets an XA Error in the prepare phase, the NodeAgent gets BAD_SERVER_ID and the EJB hosting JVM on the nodeagent's node gets a Timeout due to no client activity for 60 seconds.
Resolution : restart the JVM that is getting the timeout due to no client activity. A PMR has been opened with IBM to fix this issue.
Reason : ReactivateServers is a method where all the JVMs on a node are requested to register their EJBs. You could find that in the nodeagent ffdc logs there is a NO_PERMISSION related to this method. This generally happens when a nodeagent is restarted and th underlying EJB hosting JVMs on that node are not restarted after that.
Saturday, December 19, 2009
Subscribe to:
Post Comments (Atom)
PM24003 was created to fix this issue, the issue is that when a nodeagent is restarted without restarting the EJB hosting App servers on the node, it may fail to re-register the IORs of the EJBs, unless the application server is recycled.
ReplyDeleteThis ifix makes sure that the security context is ignored and the JVM is allowed to register its EJBs to the nodeagent, thereby allowing the recycling of a nodeagent without having to recycle all EJB hosting app servers on that node.
POst a comment and I will get back if you need more info. You will get a reactivateServers error in your nodeagent ffdc log, could be a CORBA Transient or a CORBA no_permission.