I have three A-MQ brokers inside Jboss Fuse 6.1 fabric, one master and two slaves. Back-end storage for kahadb is NFSv4. Now we have issues on the storage side and the nodes are loosing the connection. In dmesg:

NFS: server my.nfs.server.com error: fileid changed
fsid 0:13: expected fileid 0x32, got 0xc000c

At some point (usually 1-2 days) the master broker is giving:
2014-09-25 13:44:52,455 | WARN | 925772421-209484 | nio | tty.io.nio.SelectChannelEndPoint 697 | 92 - org.eclipse.jetty.aggregate.jetty-all-server - 8.1.14.v20131031 | handle failed
java.lang.OutOfMemoryError: GC overhead limit exceeded

And the slave cannot pick-up because the master don't release the NFS lock. Then I have to manually stop the master container.
Note that this server has very low load, 50 messages per day or something. And nagios checks to jolokia every few minutes.