I realised that the previous example I gave is flawed. Setting the max average to 10ms means that test can reasonably fail once the total elapsed time exceeds 1000ms (i.e. 100 x 10ms), given that it can't possibly succeed, even if the remaining iterations take no time at all. However, it was one of many attempts I'd made, most of which were more sensible.
So, here's a concrete example of something I've just tried.
I have a test step called Create Order, with a Step Average assertion, as follows:
Minimum requests: 80
Max average: 500
Max errors: 0
Sample interval: 20
My load test is set to Threads:1, Strategy: Simple, Test Delay:0, Random: 0.0, Limit: 100 total runs.
It failed on my assertion after two iterations:
Average [760] exceeds limit [500] [threadIndex=0]
The stats grid for the Create Order step shows:
Min: 372, Max: 760, Avg: 566, Last: 372, Cnt: 2
So, it's done two iterations, taking 372 and 760ms, giving an average of 566ms. But why did the assertion fire before the 80 minimum requests was reached? The tooltip for Minimum Requests shows "Minimum number of steps before asserting". Even if that meant total steps (rather than the step I'm asserting on), my test case only has 14 steps, so 2x14 is nowhere near 80.
I'd be very grateful if someone can point out what I'm doing wrong, as this is making the whole approach unworkable at the moment.
Thanks,
Ben.
