Failover client exceeding maximum connections

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Failover client exceeding maximum connections

Marc Boorshtein
I'm using org.apache.activemq.ActiveMQSslConnectionFactory with the url
"failover:(ssl://host:port)?initialRreconnectDelay=10000" and am seeing two
odd behaviors:

1.  Every few minutes we're seeing the client re-establish connections
2.  Eventually the server exceeds its connections and stops accepting new
connections

Both the client and server are running in containers in Kubernetes
communicating via a k8s service.  Both are running amq 5.15.9.  We have to
kill the amq pod and restart it to get everything working again once we
exceed the number of connections.

Any pointers on how to debug and alleviate this issue would be appreciated.

Thanks
Marrc
Reply | Threaded
Open this post in threaded view
|

Re: Failover client exceeding maximum connections

Tim Bain
Are you saying that the client is re-establishing its connections and the
old connections are staying connected (forever)? Do both the client and the
broker show the old connection as active, or does one side believe it's
closed while the other thinks it's still open?

Is there any indication in either the client logs or the broker logs that
one of them is initiating a disconnection?

Tim

On Thu, Oct 3, 2019, 1:32 PM Marc Boorshtein <[hidden email]> wrote:

> I'm using org.apache.activemq.ActiveMQSslConnectionFactory with the url
> "failover:(ssl://host:port)?initialRreconnectDelay=10000" and am seeing two
> odd behaviors:
>
> 1.  Every few minutes we're seeing the client re-establish connections
> 2.  Eventually the server exceeds its connections and stops accepting new
> connections
>
> Both the client and server are running in containers in Kubernetes
> communicating via a k8s service.  Both are running amq 5.15.9.  We have to
> kill the amq pod and restart it to get everything working again once we
> exceed the number of connections.
>
> Any pointers on how to debug and alleviate this issue would be appreciated.
>
> Thanks
> Marrc
>