Loreno Oliveira [
http://community.jboss.org/people/lorenooliveira] created the discussion
"Re: JBoss memory usage pattern"
To view the discussion, visit:
http://community.jboss.org/message/589093#589093
--------------------------------------------------------------
I tried with the JVM flag, and I'm getting the same measurements. The only difference
is the time between collections. When I just watch the output of verbosegc, it takes
longer to collect the garbage. When a both watch the output of verbosegc and also monitor
through visualvm, the collections are much more frequent, but the usage is kept in the
same window.
I can say that the CPU usage is low. This is a big machine, with a lot of processors and
6gb dedicated to this vm. CPU utilization is under 10%, with some short peaks of 30%.
When this snapshot was taken, I can say that 95% of the normal workload was not present,
but there where at least one more monitoring tool pooling data from the application
(zabbix).
As it is a production server, I'm afraid of trying a memory dump. But I took two
histograms of the server, one with low heap usage, and the other close to the top of the
heap usage. The difference between them was char arrays, Strings, int arrays, and byte
arrays. Because they are histograms, I cannot track back who is instantiating these
objects. However, doing this in the same application running in the development
environment and with a quite lower workload, I can see that all this stuff is being
generated by the container, and not by my application...
--------------------------------------------------------------
Reply to this message by going to Community
[
http://community.jboss.org/message/589093#589093]
Start a new discussion in Performance Tuning at Community
[
http://community.jboss.org/choose-container!input.jspa?contentType=1&...]