Ramping-arrival-rate

Good day. Creating a test for my service i have some kind of trouble with setting up ramping-arrival-rate test.
As you can see in picture i always have the same trouble with harsh VU allocation from 10 or 25 to 100 in test, meanwhile i need a soft VU growth simultaneously with RPMs.


What should i do in order to achieve above mentioned?
Current test looks like:

export const options = {
scenarios: {
rar: {
executor: ‘ramping-arrival-rate’,
startRate: 5,
timeUnit: ‘60s’,
preAllocatedVUs: 110,
stages: [
{ duration: ‘10m’, target: 10000,},
{ duration: ‘10m’, target: 10000,},
{ duration: ‘10m’, target: 20000,},
{ duration: ‘10m’, target: 20000,},
{ duration: ‘10m’, target: 30000,},
{ duration: ‘10m’, target: 30000,},
{ duration: ‘10m’, target: 40000,},
{ duration: ‘10m’, target: 40000,},
{ duration: ‘10m’, target: 50000,},
{ duration: ‘10m’, target: 50000,},
],
},
},
};

Hi @psixoz1,

I am not certain what you mean by “harsh” and “soft” here.

But arrival-rate executors meant for you to say how much iterations should be started and how many VUs should be able to do that. Not actually caring about which VU will be running or how many VUs will be running at the same time.

You are seing hte VUs increase as iterations can only be executed if there is a free VU and 1 VU can only execute one iteration. So if you have 100 iterations currently going on 100 VUs and by your arrival rate plan you need to start 1 more iterations - you need to use 101 VUs now.

I expect the spike you see to be because the SUT started returning slower and slower (you can see the http_req/s platoing just before that) which means that there are now more http_reqs ongoing keeping VUs active, meaning more VUs will need to be added.

Looking at the graph and your configuration something doesn’t add up thouhg, as you should get to 10k req/m and then keep that- but you keep going up for all that time, so maybe you are not running exactly this configuration?

I am not certain what you mean by “harsh” and “soft” here

just look at a pic. You can see the bunch of generated VUs likely in a minute from to 25 to 100.
I need a test where VUs are generated as slowly crawling proccess. how can i write test params to achieve that?
“http_req/s platoing just before that” → that was all ok. i showed the full test stages. its just a ladder