[jboss-jira] [JBoss JIRA] Created: (JGRP-520) OOM with RpcDispatcher

[Bela Ban (JIRA)]

OOM with RpcDispatcher
----------------------

                 Key: JGRP-520
                 URL: http://jira.jboss.com/jira/browse/JGRP-520
             Project: JGroups
          Issue Type: Bug
         Environment: 1.5.0_10-b03;Sun Microsystems Inc.;mixed mode, sharing;Windows;32 bit JVM
Windows Server 2003 Standard x64
            Reporter: Bela Ban
         Assigned To: Bela Ban
            Priority: Critical
             Fix For: 2.5
         Attachments: SimpleClusterManager.java

[from Patrick Eger]

Attempting to isolate the problem, I ran into a (probably unrelated)
other OutOfMemoryError problem with JGroups 1.5b2. Got an
OutOfMemoryError after ~3 minutes of running the attached
TestJGroups.java, 3 instances in parallel. Also attached is the stack
config and a couple YJP views of the HPROF dump as generated by
-XX:+HeapDumpOnOutOfMemoryError.

The test is looping single-threaded, calling a remote method via
RpcDispatcher every 100 ms which returns a few KB String. Looks like all
the memory is retained instances of org.jgroups.stack.IpAddress
(~1,000,000 for a ~3 minute run (seems like a lot?)), held by
pbcast.NAKACK.sent_msgs. I have the HPROF dump if you need, its about 15
megs zipped so I assumed would not make it through the mail filters.

Is this a real OOME or is it a stack configuration/tuning thing? I have
used the default UDP config, but with increased thread_pool and
oob_thread_pool sizes.

Options
-----------------------------------------
-agentlib:yjpagent
-server
-Xmx64m
-XX:+HeapDumpOnOutOfMemoryError
-Dcom.sun.management.jmxremote
-Dpeer_cluster_manager.XXXTestClusterXXX.jgroups_config=C:\tmp\testcfg.x
ml

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://jira.jboss.com/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira