Hi,
We have some beefy machins as injectors (2 Real CPU, 4 Logical with 6Gig of
RAM). To sqeeze more out of them we are running multiple Jmeters from the
same machine using one Jmeter on the same machine to start/stop the other
JVM's.
In our tests we use a lot of beanshell's that start up 4 threads per client
keeping the connections active for the life time of the test.
All looked good until we hit around 2200 threads (1200 per JVM) after which
our logins carried on working but some of the System.out messages from the
JVM's stopped and when the system shutdown the threads it did not log out
all the objects cleanly with clients still left connected to the server. (At
the end of the run we had 3000 confirmed live connections to the server from
the Jmeter machine).
I know the total number of threads is a lot but we are running 2.6 kernel
and the CPU was always > 90% idle (At this point the threads do not do much
after login). The only load on the system seemed to be RMI stuff with the lo
interface running at around 700k/sec.
There was no out of memory messages from the Heap, I have not checked GC
stats but pauses caused by the GC should not show odd behaviour like this...
Has anyone else seen anything like this?
Also we are splitting the Jmeter into several JVM's as I had read on this
forum that 1500 Threads is about the max one JMeter can handle, is this
correct?
Cheers
R
_________________________________________________________________
Download the new Windows Live Toolbar, including Desktop search!
http://toolbar.live.com/?mkt=en-gb
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
