I would attach a JMX console to the broker and look at the broker's
memory percentage used attribute. If it's 100% then that's the reason
that the broker is blocked. This memory used has nothing to do with
the JVM memory configuration BTW. It's a configurable setting in the
broker XML and can be changed on the fly using the jmx console by
setting the memory limit attribute on the broker JMX object.
On 6/21/06, Igor Bogicevic <[EMAIL PROTECTED]> wrote:
I actually have one more detail that might be interesting... During
some experiments I've when the queue is blocked I received this output
from bstat:
ACTIVEMQ_HOME: /data1/activemq/incubator-activemq-4.0.1
ERROR: java.lang.RuntimeException: Failed to execute query task.
Reason: java.io.IOException: Failed to retrieve RMIServer stub:
javax.naming.ServiceUnavailableException [Root exception is
java.rmi.ConnectExc
eption: Connection refused to host: localhost; nested exception is:
java.net.ConnectException: Connection refused]
ERROR: java.lang.Exception: java.io.IOException: Failed to retrieve
RMIServer stub: javax.naming.ServiceUnavailableException [Root
exception is java.rmi.ConnectException: Connection refused to host:
localhost
; nested exception is:
java.net.ConnectException: Connection refused]
also, tests are performed on 3 different boxes on same switch
(full-duplex 1Gbit) and producers, consumers and broker are on
separate boxes with plenty of memory (-Xmx8192M -Xms1024M for broker),
with no visible trail of memory lack or I/O blocking on the each of
the boxes).
Thanks,
Igor
On 6/21/06, Igor Bogicevic <[EMAIL PROTECTED]> wrote:
> Sorry, I forgot to specify the details, it's incubator-activemq-4.0
> and i am using Sun JDK 1.5_06
>
> Thanks,
> Igor
>
> On 6/21/06, James Strachan <[EMAIL PROTECTED]> wrote:
> > Which version of ActiveMQ are you using?
> >
> > On 6/21/06, Igor Bogicevic <[EMAIL PROTECTED]> wrote:
> > > Hi all,
> > >
> > > I've been testing ActiveMQ under relatively high load and number of
> > > producers and consumers (50-100 on each side) and I've attached the
> > > code that I have been using for test purposes (it's not really clean
> > > code since it's used for testing purposes only). What happens is when
> > > I launch about 50-60 consumers/producers (they run on a different
> > > boxes) and I start to stop consumer processes, producers tend to block
> > > for a shorter period, if I repeat this process for a while I get to
> > > state where's a whole queue blocked for 1-2h or longer (sometimes for
> > > good).
> > > Is this general performance problem with ActiveMQ or am I doing something
wrong?
> > >
> > > Thanks,
> > > Igor
> > >
> > >
> > >
> >
> >
> > --
> >
> > James
> > -------
> > http://radio.weblogs.com/0112098/
> >
>
--
Regards,
Hiram
Blog: http://hiramchirino.com
