Allow for early termination and partial data collection about trombi

[mhjort]

If tests are running far longer than specified that's a definitely a bug. Do you have an example that is reproducible? I could take a look on that.

[mhjort]

Early termination is a good idea. However, I have to think about it a bit so that it fits to current internal design.

[sarudak]

I think there's actually two issues here. One is (I'm guessing) that regardless of duration already started scenarios will complete so if I try to do a 30 sec test and my scenario is 5 min it takes that long no matter what and at the end of a 20 minute test the stragglers will finish their scenarios. I'm guessing this is WAD.

The other is when I push the concurrent users over 100 to say 150 or 200 then I seem to get less traffic than at 100 users and the scenarios never seem to finish even if I wait an extra 30 min past time. Not sure what's going on there. Maybe my computer just can't handle it but it's not giving me a lot of visibility into what's happening.

[mhjort]

You're right. clj-gatling checks whether it should finish after running scenario. So it will finish up all the scenarios that have been started before. If you have a small timeout value for scenario steps that's usually not a problem. Also that way ongoing requests are never interrupted (meaning that all responses are checked). For me that has been important thing but I understand that's not always needed. How long timeout you are using in your tests?

[sarudak]

I'm using standard clj-http timeouts. It's less the timeout than that I'm running 20+ steps in the scenario with a sleep-before of 1-10 sec each to simulate user delay. Cutting the delay for my testing helps a lot so I made that configurable.

[mhjort]

Implementing stop condition check after each scenario step is quite trivial. I added a separate issue about this #24 .

[mhjort]

Solved in #24