to many active users when using constantUsersPerSec

Gatling (Open-Source)
1

Hi,

I noticed strange behaviour when using constantUsersPerSec. For scenario configuration:

constantUsersPerSec(130) during (4 minutes)

active users count has some unexpected peak at the beginning of the test:

I changed the configuration to something like this:

.inject(

constantUsersPerSec(30) during (30 seconds),

constantUsersPerSec(60) during (30 seconds),

constantUsersPerSec(90) during (30 seconds),

constantUsersPerSec(120) during (30 seconds),

constantUsersPerSec(150) during (30 seconds)

)

And the same thing happens:

How to explain such behaviour?

2

Because you mostly likely suffer from a huge response time spike that causes virtual users lifespan to increase as much.
As you use an open workload model, new virtual users keep on arriving while existing users took way longer to die, so hence the active users bump.

3

Yeah, just wanted to be sure that this is SUT fault, I cannot spot anything unusual, because after the peak, everything returns to normal… profiling, logging, monitoring also looks ok…

W dniu poniedziałek, 14 maja 2018 14:53:21 UTC+2 użytkownik Stéphane Landelle napisał: