Resolution:
Description:
============
BEMM application memory usage increases gradually and crashes after a few hours.


Environment:
===========
TIBCO BusinessEvents 4.0.1 GA
TIBCO BusinessEvents 4.0.1 HF1


Symptoms:
========
By monitoring the BEMM process, it can be seen that the memory foot print increases gradually and the process crashes after running few hours.


Cause:
=====

1). All monitored agents send update events to the BE-MM server .

2). These events are accumulated by the internal query agent in BE-MM server instance.

3). The results of the query agent are sent via a local channel to the inference agent in BE-MM in server instance.

4). The inference agent updates the charts.

5). The monitored agents send too many events to the BE-MM server which eventually causes the local channel to get flooded. This causes out of memory errors in the BE-MM server instance.



Resolution:
==========

Apply BE 4.0.1 HF2 or higher release and control the event publishing rate of the monitored agents. This can be achieved by:

1). Adding com.tibco.be.metric.publisfh.interval (default is 10000 msecs) and com.tibco.be.metric.resolution (default is 5000 msecs) to the CDD of the monitored agents , NOT BE-MM cdd.

2). Suggested values are  300000 for both com.tibco.be.metric.publish.interval and com.tibco.be.metric.resolution.

3). Adding to BE-MM cdd will not make a difference since the monitored agents will keep publishing data at their default rate.