loosing connection after some time

[jmesnil]

loosing connection after some time

[zygm0nt]

Tried to check if there is a time correlation for those error messages.

Tue Nov 12 2013 12:53:14 GMT+0100 (CET): <<< MESSAGE message-id:ID:correspondence-55764-1384256136182-11:1:1:1:1 destination:/topic/sales timestamp:1384257194511 expires:0 subscription:sub-0 priority:0 sdfdf

Tue Nov 12 2013 12:58:15 GMT+0100 (CET): Whoops! Lost connection to ws://localhost:61614/

It seems the connection is lost after 5 minutes of inactivity.

Is that true? Can I disable this check? I would like to have sessions that are based on heartbeat and not some other inactivity checks (if-any).

[jmesnil]

Yes, it looks like there is an issue with the heart-beating check.

Which version of ActiveMQ are you using? There is a related bug in 5.8.0.

If you enabled the logs with stomp.js, what values do you get for the heart-beat header in the CONNECTand CONNECTED frames?

[zygm0nt]

I'm using ActiveMQ 5.9.0.

The logs are like this:

 >>> CONNECT login:abc passcode:abc accept-version:1.1,1.0 heart-beat:10000,10000
<<< CONNECTED heart-beat:0,0 session:ID:correspondence-55764-1384256136182-2:4 server:ActiveMQ/5.9.0 version:1.1

Is heart-beat:0,0 a wrong return header from activemq?

[jmesnil]

yes, you are hitting the bug in activemq https://issues.apache.org/jira/browse/AMQ-4740 that is targeted for 5.10.0.

Until you upgrade ActiveMQ, the best way is to disable heart-beating from stomp.js:

 client.heartbeat.outgoing = 0;
 client.heartbeat.incoming = 0;

[zygm0nt]

Thanks for quick help! I'll stick with this solution.

[psharpNumerex]

while we're all waiting for 5.10, I tried setting the heartbeat values to 0 and still get disconnected after 5 minutes. Is there something else I can do to prevent this? I'm reconnecting after the loss, but it's crazy annoying and I've seen some erratic behavior on chrome.

[jmesnil]

Unfortunately, there is nothing you can do on stomp.js side if ActiveMQ disconnects you. Until they fix their bug, you will have to handle this case and reconnect every time.

[chinshaw]

I have been investigating this problem and think I have a work around that will not disconnect you. It seems that ActiveMQ 5.9, 5.10 has an issue with handling the heartbeat event after about the thrid PING. But in 5.10+ a the StompInactivityMonitor was added which will handle the PONG/ server side keep alive. If you configure the connection so that only the server will handle the keep alive it appears not to disconnect the session. I have had my connection going for over 20 minutes while using the server side heartbeat. Quick note though I am using a development version of ActiveMQ, but I tested prior to see if the outgoing heartbeat worked but it was the same problem as 5.9 and 5.10 (Disconnection after a couple of tries, maybe 20 seconds). I did also add the option to use 1.2 and it is connected at 1.2 version of stomp.

client.heartbeat.outgoing = 0;
client.heartbeat.incoming = 20000;

">>> CONNECT
accept-version:1.2,1.1,1.0
heart-beat:0,20000

�" ng.html:145
"<<< PONG" ng.html:145
"<<< CONNECTED
server:ActiveMQ/5.11-SNAPSHOT
heart-beat:20000,0
session:ID:hostname
version:1.2