High volume LaadUI => SoapUI test seems to break off threads
LoadUI 1.0.1/SoapUI 3.6.1
The issue we are seeing is somewhat strange:
Our LoadUI project contains a SoapUI Test Runner that invokes a TestCase in SoapUI with several test steps. We have added debug statements in the Test Steps by creating Script Assertions that do a log.info "test step X" as well as a Script Test Step that does the same.
When we increase the load (higher volume from the Fixed Rate Generator, for example from 2 to 5,10 or 15 per second. with an average execution time per Test Case of about 350 ms), we have the impression that even though we have no time out failures or SOAP Assertion Failures, not all test cases are completed successfully. More specifically, we see a behavior (thanks to our logging statements) similar to this:
Step A - 100%
Step B - 98%
Step C - 96%
Step D - 95%
That is: even though no failures are reported, some of our test cases abort before they have executed all test steps! It seemed that when we took out one step (that was by far the slowest), the behavior was far less pronounced.
Has anyone experienced similar behavior? Does anyone have an explanation for such behavior?
Thanks for any insights,
Lucas
The issue we are seeing is somewhat strange:
Our LoadUI project contains a SoapUI Test Runner that invokes a TestCase in SoapUI with several test steps. We have added debug statements in the Test Steps by creating Script Assertions that do a log.info "test step X" as well as a Script Test Step that does the same.
When we increase the load (higher volume from the Fixed Rate Generator, for example from 2 to 5,10 or 15 per second. with an average execution time per Test Case of about 350 ms), we have the impression that even though we have no time out failures or SOAP Assertion Failures, not all test cases are completed successfully. More specifically, we see a behavior (thanks to our logging statements) similar to this:
Step A - 100%
Step B - 98%
Step C - 96%
Step D - 95%
That is: even though no failures are reported, some of our test cases abort before they have executed all test steps! It seemed that when we took out one step (that was by far the slowest), the behavior was far less pronounced.
Has anyone experienced similar behavior? Does anyone have an explanation for such behavior?
Thanks for any insights,
Lucas
