Httpbind responds after 'wait' value

Openfire Openfire Dev
1

I’m connecting to openfire http-bind through the browser using jabberwerx.

The “wait” value of the BOSH connection is 30, however what I’m noticing is that after several empty longpoll requests (3-5 minutes) there will be one longpoll request that openfire responds to with an empty request (which is normal) but which exceeds the ‘wait’ value of 30 seconds (definitely not normal).

With debug logging enabled, I see the following:

Fri Jan 27 12:56:29 EST 2012: HTTP RECV(f5540e93):

Fri Jan 27 12:56:59 EST 2012: HTTP SENT(f5540e93):

Fri Jan 27 12:56:59 EST 2012: HTTP RECV(f5540e93):

Fri Jan 27 12:57:37 EST 2012: HTTP SENT(f5540e93):

Fri Jan 27 12:57:37 EST 2012: HTTP RECV(f5540e93):

Fri Jan 27 12:57:45 EST 2012: HTTP RECV(f5540e93):

Fri Jan 27 12:57:45 EST 2012: HTTP SENT(f5540e93):

Fri Jan 27 12:57:45 EST 2012: HTTP RECV(f5540e93):

Fri Jan 27 12:57:45 EST 2012: HTTP ERR(f5540e93): terminate, policy-violation.

Instead of responding with an empty response after 30 seconds, for request with rid=1781211427, openfire responds after 38 seconds.

Has anyone encountered such an issue with the Jetty Continuations such that openfire sends a response way after the “wait” value has expired?

Thanks,

Conrad

2

As a follow on (and separate question from above), because this one longpoll request ends up taking longer than the “wait” value of 30 seconds, in the browser, something in jabberwerx chokes up and it ends up sending two requests with the same rid which ends up with a policy-violation response and openfire kills the session.

I tried turning on the property for “ignoreOveractivity” and that does the trick, but after it recovers from the two requests with the same rid, openfire starts responding to longpoll requests immediately, even though there is no event to pass back in the response. And as a result, the browser starts getting a flood of empty requests and responses.

Has anyone seen this behavior?