With EAP 6.3.0.DR2, a single web application deployed (both with distributable and non-distributable versions), several requests made to it, one can get ShutdownChannelGroupException on a clean shutdown, for a no apparent reason. I believe this is a random race condition, yet it is noteworthy that it occurred with mod_cluster and JGroups subsystems in place, standalone-ha profile. {noformat} (Thread-105) Exception in thread "Thread-105" java.nio.channels.ShutdownChannelGroupException (Thread-105) at sun.nio.ch.Invoker.invokeIndirectly(Invoker.java:210) (Thread-105) at sun.nio.ch.UnixAsynchronousSocketChannelImpl.finishRead(UnixAsynchronousSocketChannelImpl.java:432) (Thread-105) at sun.nio.ch.UnixAsynchronousSocketChannelImpl.finish(UnixAsynchronousSocketChannelImpl.java:191) (Thread-105) at sun.nio.ch.UnixAsynchronousSocketChannelImpl.onEvent(UnixAsynchronousSocketChannelImpl.java:213) (Thread-105) at sun.nio.ch.EPollPort$EventHandlerTask.run(EPollPort.java:293) (Thread-105) at java.lang.Thread.run(Thread.java:744) {noformat} Environment is: java version "1.7.0_51" OpenJDK Runtime Environment (rhel-2.4.4.1.el6_5-x86_64 u51-b02) & OpenJDK Runtime Environment (rhel-2.4.4.1.el7-x86_64 u51-b02) Any ideas?
It might actually not be limited to JBossWeb, I guess JGroup use NIO as well, so mabye lThon might have seen it somewhere as well... WDYT lThon? :-)
I don't remember seeing this. It's not clear if this happens in clustering (I guess you get this in a clustered setting) or in the web subsystem. Can you [try to] reproduce this in the standalone profile (i.e., no "ha")?
AFAIK, this is specific to standalone-ha (ha) profile, inherently to full-ha (not tested). On the other hand, the fact that I didn't see it proves nothing, provided it's some kind of a race condition. I will bring this email from the internal jboss-support-eap6 list to your attention: [01035878] Http11NioProtocol not being able to handle concurrent requests Aaron? WDYT? Is it related?
*** This bug has been marked as a duplicate of bug 1075695 ***