Perhaps not relevant, but I'm using a python client with a C++ broker with the (QMF) heartbeat interval (configured in the broker) at ten seconds. I had occasion to instrument the actual time between heartbeats, and while usually very close to (but always longer than) ten seconds, I have seen odd single instances that are significantly longer, one as long as 16.4 seconds (presumably the broker was "busy" with something else). This was in a CentOS VM running under vmware player on a Windows box, so my variability may be worse than usual. But if my level of variability do occur more generally, I wonder if there may be a bug related to setting the heartbeat to intervals shorter then the variability.
Bill On Fri, May 2, 2014 at 5:02 PM, Chuck Rolke <[email protected]> wrote: > Some more information would be helpful to discover your problem: > > Are you running a specific release or a private build? > Configuration details? (standalone broker, federated, clustered) > How are the brokers started? > Message pattern hints: large messages, small messages, idle? > Native C++ clients or language binding? > Is a reproducer available? > > -Chuck > > ----- Original Message ----- > > From: "trivedi_ravi13" <[email protected]> > > To: [email protected] > > Sent: Friday, May 2, 2014 3:05:19 PM > > Subject: Qpid client crashes in heartbeat operation > > > > I have almost 15-20 clients running on my system. all clients use > heartbeat > > to detect connection failure. these clients started to get crashed > randomly > > after enabling heartbeat. Timeout is set to 2 sec for some app and 5 sec > for > > others. Below is attached core dump output. It appears to be getting > > terminated in qpid::sys:Timer::run() method. > > > > > > (gdb) bt full > > #0 0x0fe76444 in raise () from /lib/libc.so.6 > > No symbol table info available. > > #1 0x0fe7b694 in abort () from /lib/libc.so.6 > > No symbol table info available. > > #2 0x0fe6d778 in __assert_fail () from /lib/libc.so.6 > > No symbol table info available. > > #3 0x0f9357d8 in qpid::sys::Timer::run() () from > > /usr/lib/libqpidcommon.so.2 > > No symbol table info available. > > #4 0x0f803984 in ?? () from /usr/lib/libqpidcommon.so.2 > > No symbol table info available. > > #5 0x0fd76d8c in ?? () from /lib/libpthread.so.0 > > No symbol table info available. > > #6 0x0ff25a90 in clone () from /lib/libc.so.6 > > No symbol table info available. > > > > > > We need this feature to detect failure as soon as possible. Is there any > > other way to do that. I believe SO_KEEPALIVE is not set in sockets in > qpid. > > > > > > > > -- > > View this message in context: > > > http://qpid.2158936.n2.nabble.com/Qpid-client-crashes-in-heartbeat-operation-tp7607861.html > > Sent from the Apache Qpid users mailing list archive at Nabble.com. > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [email protected] > > For additional commands, e-mail: [email protected] > > > > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: [email protected] > For additional commands, e-mail: [email protected] > >
