Gatling flavor: scala
Gatling build tool: maven
Trying to execute Gatling performance tests with 20 and 40 users.
But user injection is not proper. 40 users test taking too much time for ramp up.
Also there are many spikes in user graph during load test.
Scenario 1:
20 users with 20 seconds ramp up
Users are ramped up correctly but there are many spikes throughput the test duration.
Scenario 2
40 users with 40 seconds ramp up
40 users are taking more than 4 minutes to rap up instead of 40 seconds.
there are many spikes throughput the test duration.
Scenario 3
AtOnceUsers 40
All users are ramped up correctly (at once) and also user count is constant throughout the test duration. But this scenario is not useful for our test executions as we need some ramp up for the test.
Why there are problems in Scenario 1 and 2.
Problem 1: 40 users are not ramped up within 40 seconds. It’s taking about 4 minutes to rap them up.
Problem 2: User count is not consistent throughput the test duration. There are a lot of spikes in user graphs.
Someone can please help.
Regards,
AP
There’s no problem with the injection profile.
It’s just that the “Active Users” metric is not what you think: Timings
Might be my understanding is incorrect for active users.
But when I analyse the results, response time for 20 users test is 840 ms. No of samples is 6836. Throughput is 7.4 TPS.
For 40 users test, response time is 822 ms, no of samples is 7899 and throughput is 8.6.
If the response times are same for both the tests and user count is double then throughput should also be around twice. But it’s not happening.
Executed 15 minutes load tests with 20 users and 40 users.
Response times are almost same for both the tests but TPS for 40 users test is less than 20 users test.
This is really confusing.
Earlier we executed same tests with JMeter and we were getting double TPS in 40 users test compared to 20 users test.
First, you’re still using an old version. We require that you first upgrade to the latest release (3.14.1 as of now) so we can try to help you.
Then, please provide code samples.
Without seeing your code, and possibly being able to run it on our side, there’s no way to tell what your injection profile and your scenario are doing and hence help you.
I observed one thing that the number of requests count is always 20, no matter how many virtual users we are running.
We’ve released Gatling 3.14.2 that replace the “Active Users” metric with “User start rate” and “Concurrent users”.
You’ll be able to see for yourself that Gatling behaves as you’ve instructed.
