You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@activemq.apache.org by Jan Stette <ja...@gmail.com> on 2007/06/25 13:55:20 UTC
ActiveMQ Dropping Connections
I'm currently trying to incorporate ActiveMQ 4.1.1 into a project (Java 6,
tested on Windows XP so far).
We're seeing a lot of cases in the logs where ActiveMQ is apparently
dropping and re-establishing connections. This is with only two JVMs,
running on a single host, "peer" transport, setting up TCP connections
between JVMs and clients in each JVM connecting to the embedded broker.. It
seems to happen when the processes are heavily loaded - even when there is
no messaging going on!
Is there some kind of watchdog thread or timeout that is causing this to
happen? How can I configure this to be less agressive in bouncing
connections?
Regards,
Jan
Re: ActiveMQ Dropping Connections
Posted by Jan Stette <ja...@gmail.com>.
I've finally got around to testing the latest ActiveMQ snapshot (from 23
July). It's exhibiting the same problem. Basically, while a server is
doing some heavy processing, so it's maxing out the CPU for a couple of
minutes, connections to this server's peer are being dropped and
re-established. This is even with very little (if any) JMS traffic
happening at the time.
The inactivity timeout setting mentioned below doesn't make a difference
either.
The logs contain repeated sequences like this:
4-Jul-2007 15:16:31
org.apache.activemq.network.DemandForwardingBridgeSupport stop
INFO: peer--42024-1185286366163-0-0 bridge to peer--50754-1185286411884-0-0
stopped
24-Jul-2007 15:16:31
org.apache.activemq.network.DiscoveryNetworkConnectoronServiceAdd
INFO: Establishing network connection between from
vm://peer--42024-1185286366163-0-0 to tcp://somehost:43053
24-Jul-2007 15:16:32
org.apache.activemq.network.DemandForwardingBridgeSupport startLocalBridge
INFO: Network connection between vm://peer--42024-1185286366163-0-0#16 and
tcp://localhost/127.0.0.1:43053(peer--50754-1185286411884-0-0) has been
established.
<...>
24-Jul-2007 15:16:44
org.apache.activemq.network.DemandForwardingBridgeSupport stop
INFO: peer--42024-1185286366163-0-0 bridge to peer--50754-1185286411884-0-0
stopped
<...>
24-Jul-2007 15:16:44
org.apache.activemq.network.DiscoveryNetworkConnectoronServiceAdd
INFO: Establishing network connection between from
vm://peer--42024-1185286366163-0-0 to tcp://somehost:43053
<...>
24-Jul-2007 15:16:44
org.apache.activemq.network.DemandForwardingBridgeSupport startLocalBridge
INFO: Network connection between vm://peer--42024-1185286366163-0-0#18 and
tcp://localhost/127.0.0.1:43053(peer--50754-1185286411884-0-0) has been
established.
Does anyone have suggestions how we can debug or work around this? My
company is evaluating ActiveMQ at the moment and it's of critical importance
that we can get it working reliably, or we'll have to look somewhere else
for a JMS implementation. :-/
Many thanks,
Jan
On 26/06/07, Jan Stette <ja...@gmail.com> wrote:
>
> Hi James, thanks for your reply.
>
> I tried the option you mention, but it didn't seem to make a difference.
> I will try the latest 5.0 snapshot as well to see if that's any better.
>
> Regards,
> Jan
>
>
> On 25/06/07, James Strachan <ja...@gmail.com> wrote:
> >
> > On 6/25/07, Jan Stette <ja...@gmail.com> wrote:
> > > I'm currently trying to incorporate ActiveMQ 4.1.1 into a project
> > (Java 6,
> > > tested on Windows XP so far).
> > >
> > > We're seeing a lot of cases in the logs where ActiveMQ is apparently
> > > dropping and re-establishing connections. This is with only two JVMs,
> > > running on a single host, "peer" transport, setting up TCP connections
> >
> > > between JVMs and clients in each JVM connecting to the embedded
> > broker.. It
> > > seems to happen when the processes are heavily loaded - even when
> > there is
> > > no messaging going on!
> > >
> > > Is there some kind of watchdog thread or timeout that is causing this
> > to
> > > happen? How can I configure this to be less agressive in bouncing
> > > connections?
> >
> > This could be a bug that was fixed in trunk...
> > http://issues.apache.org/activemq/browse/AMQ-1146
> >
> > a workaround is to increase maxInactivityDuration
> > http://activemq.apache.org/configuring-wire-formats.html
> >
> > --
> > James
> > -------
> > http://macstrac.blogspot.com/
> >
>
>
Re: ActiveMQ Dropping Connections
Posted by Jan Stette <ja...@gmail.com>.
Hi James, thanks for your reply.
I tried the option you mention, but it didn't seem to make a difference. I
will try the latest 5.0 snapshot as well to see if that's any better.
Regards,
Jan
On 25/06/07, James Strachan <ja...@gmail.com> wrote:
>
> On 6/25/07, Jan Stette <ja...@gmail.com> wrote:
> > I'm currently trying to incorporate ActiveMQ 4.1.1 into a project (Java
> 6,
> > tested on Windows XP so far).
> >
> > We're seeing a lot of cases in the logs where ActiveMQ is apparently
> > dropping and re-establishing connections. This is with only two JVMs,
> > running on a single host, "peer" transport, setting up TCP connections
> > between JVMs and clients in each JVM connecting to the embedded
> broker.. It
> > seems to happen when the processes are heavily loaded - even when there
> is
> > no messaging going on!
> >
> > Is there some kind of watchdog thread or timeout that is causing this to
> > happen? How can I configure this to be less agressive in bouncing
> > connections?
>
> This could be a bug that was fixed in trunk...
> http://issues.apache.org/activemq/browse/AMQ-1146
>
> a workaround is to increase maxInactivityDuration
> http://activemq.apache.org/configuring-wire-formats.html
>
> --
> James
> -------
> http://macstrac.blogspot.com/
>
Re: ActiveMQ Dropping Connections
Posted by James Strachan <ja...@gmail.com>.
On 6/25/07, Jan Stette <ja...@gmail.com> wrote:
> I'm currently trying to incorporate ActiveMQ 4.1.1 into a project (Java 6,
> tested on Windows XP so far).
>
> We're seeing a lot of cases in the logs where ActiveMQ is apparently
> dropping and re-establishing connections. This is with only two JVMs,
> running on a single host, "peer" transport, setting up TCP connections
> between JVMs and clients in each JVM connecting to the embedded broker.. It
> seems to happen when the processes are heavily loaded - even when there is
> no messaging going on!
>
> Is there some kind of watchdog thread or timeout that is causing this to
> happen? How can I configure this to be less agressive in bouncing
> connections?
This could be a bug that was fixed in trunk...
http://issues.apache.org/activemq/browse/AMQ-1146
a workaround is to increase maxInactivityDuration
http://activemq.apache.org/configuring-wire-formats.html
--
James
-------
http://macstrac.blogspot.com/