It was noticed that all of the sesame (system agent) processes running in the North grid test environment were not publishing updates. If qpid-tool is run against the broker (North-15), no sysimage objects are seen even though there are several sesame daemons running. The pstacks of some of the sesame processes were observed and found to be in normal "idle" state. More investigation is needed.
A potential workaround has been committed upstream at revision 727485. This won't solve the root problem, but if the qmf-agent session is lost for any reason, this change will cycle the connection for a fresh start. It will also log the exception that caused the problem to begin with. I am bumping the target to 1.1.1 and not setting to MODIFIED in hopes that more information will emerge leading to a solution to the root cause.
BZ 477036 is likely related to this bug. I think that the symptom described in that bug is the cause for session loss in the qmf agents. In other words, messages sent from an agent to the "qpid.management" topic exchange are dropped with a session-ending exception when there is a backed-up queue bound to the exchange.
The committed workaround appears to have solved the problem.
Not reproducable on old version. After creating a queue bound to exchange qpid.management and filling it up to its limit, the connection to e.g. cumin is not lost and I do not see any Exception in the logs. Issue not seen on current errata versions of qpidd-0.5.752581-3.el5 qmf-0.5.752581-3.el5
An advisory has been issued which should help the problem described in this bug report. This report is therefore being closed with a resolution of ERRATA. For more information on therefore solution and/or where to find the updated files, please follow the link below. You may reopen this bug report if the solution does not work for you. http://rhn.redhat.com/errata/RHEA-2009-0434.html