from a quick look at one of those stack traces, it looks like a thread is blocked on a socket write, it may be that the socket is half closed and that the write needs to timeout at the tcp level. that can take some time.
There is a writeTimeout filter that can help at the activemq level. Have a peek at https://issues.apache.org/jira/browse/AMQ-1993 On 30 October 2012 21:31, James Green <james.mk.gr...@gmail.com> wrote: > This is happening repeatedly causing memory consumption to balloon locally > and message queues to build remotely. > > Googling around isn't revealing much. Can anyone take a look and tell me if > this stuff is normal or not? Need to narrow down the investigations. > > Thanks, > > James > > On 30 October 2012 11:32, James Green <james.mk.gr...@gmail.com> wrote: > >> Hub with three spokes. The network between the hub and one of these spokes >> seems to be dying irregularly yet frequently. >> >> I managed to get a thread dump of the hub's ssl connector to this spoke >> along with the thread it was waiting on: >> >> http://pastebin.com/ZjeSxDK7 >> >> I also captched a full thread dump on the hub: >> >> http://pastebin.com/Qu113JhQ >> >> On the spoke the thread dump button was disabled but I captured the >> equivalent individual threads: >> >> http://pastebin.com/1hwb1BDp >> >> FWIW we just restarted the spoke and the network connector appears to be >> stuck once again. Not sure what we can actually do now... >> >> Can anyone offer advice, perhaps a JIRA issue? This is with 5.7.0. >> >> James >> >> -- http://redhat.com http://blog.garytully.com
