Hi sebb,
As your test plan is calling the database via the web application, in
theory enough users of the application can cause the same problem.
You have proved that the web application can bring down the database
(which is useful to know).
Whether it is possible for this to occur when not using JMeter depends
on how close the test plan load is to what the application is intended
to support.
Use application logs to gather stats on the page requests made by
actual users, and compare those with the load generated by JMeter.
Adjust the JMeter test plan as necessary to reflect the target load.
You also need to look at the web application to see if some of the SQL
can be changed to reduce the database load.
Actually, what I did was taking REAL data from the logs in order to
replay these with jmeter. That works fine as long as you edit this data
a little (passwords etc.). The test plan is thus as close as I might
possibly get to a real user's actions. Four csv files for four different
options to get from a log in to the check out. These four ways are then
executed by the threads in parallel.
However, what I do not know is how the actual live system reacts because
I can only test on a less performant development system. Hopefully, the
live system is better configured.
rgds
Karsten
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
