ActiveMQ disconnects and reconnects connections from time to time

[Wojciech Trzaski <Wo...@sapiens.com>]

Hi,

This is my first email using Apache mailing lists. And... I have a problem with understanding one ActiveMQ's behaviors. I don't know if it's a problem with connectivity or environment - or it is by design.

I am using ActiveMQ 5.15.10. Small cluster of servers (just 5), one main MQ broker (plus a "spare" one). Configuration is mostly the default settings, some of the timeouts are shorter and that's it. The keepAlive is set to true. ActiveMQ is up and running, but recently we noticed, that from time to time - few times a day - I can see in the logs of the broker:

Stopping connection: tcp://10.220.0.42:52367
Stopping transport tcp:///10.220.0.42:52367@61616
Closed socket Socket[addr=/10.220.0.42,port=52367,localport=61616]
Forcing shutdown of ExecutorService: java.util.concurrent.ThreadPoolExecutor@6f48400a[Running<mailto:java.util.concurrent.ThreadPoolExecutor@6f48400a[Running>, pool size = 1, active threads = 0, queued tasks = 0, completed tasks = 1]

And a minute or two later:

Setting up new connection id: ID:(...) , address: tcp://10.220.0.41:54981

Then I am getting logs about adding new consumer, publishing new tcp://(...):61616 for broker transport URI etc.

And I would not care about it (or I wouldn't even notice it) if not one problem: the message that is being sent in the meantime is delivered 30 seconds later - after the wireFormat.maxInactivityDurationInitalDelay=30000 time has passed (and after it I can see "Setting up new connection id" etc). When I decreased this time to 10000, then the latency for this single message was decreased to 10 seconds.

What can cause such behavior? Why ActiveMQ stops the connection, that is later recreated - with delay set in maxInactivityDurationInitalDelay?

Regards,

Wojciech Trząski, Java Developer