L
lecuyer
Guest
To narrow this down a bit, I would suggest analyzing the application requirements of starting new agents (especially startup procedures), and what your usage patterns are regarding starting agents 'on demand. That is, do you employ an relative static agent startup policy (e.g. prestart all/most of your agents and keep them running) or is it more dynamic (e.g. prestart few agents, and use an agressive auto-trim property to terminate agents that are not immediately needed). There is no right or wrong configuration in this regards, and it might be possible to tune this a bit. That being said, I should probably ask whether these values for rqWait and rqDuration are typical, or are they anomalous, since your average times are very respectable. How long had this broker been running when this data was collected? One of your agents has been running at least 13 days. Unfortunately the statistics presented reflect behaviour since the time the broker was started, and long running times can tend to mask the effects of specific events. Restarting you broker more frequently might yield some information. You could also consider gathering the asbman query data over time in regular intervals (e.g. every 30 seconds) using an automated script. This might help identify when these long running requests occurred. Again, you might want to consider using OEManagement to gather this level of detail.
Continue reading...
Continue reading...