JMeter: How can I keep 1000 or "X" users login requests session alive using JMeter?

Question

Here is my Test Scenario:

First, I am generating a load which includes only Login Requests (say for 1000 users). I am using "Ultimate Thread Group" and "Constant Throughput Timer" in my script. Constant Throughput Timer Value=120/sec is used. I want to run this test for five or ten minutes. My load would be held for 3 minutes.

During this Test(while loads are held), I am sending another set of login requests from the different machine (say for 100 users). I want to measure the response time of this 100 users login which I've generated from the different machine.

But My Requirement is: while I'm sending 100 users login requests, My Previous 1000 users login requests session should be alive on the server. I've also checked "Use KeepAlive" in my login sampler.

So, How can I achieve this?? How can I be sure that my previous all the login requests sessions are held on the server?

Answer 1

Although Constant Throughput Timer accepts value in Requests/min and you are saying that you will use 120/sec, so you will have to provide a value of 7200 in Constant Throughput Timer to achieve 120/sec.

Assuming there is only one request (sampler) in your Thread group and you are using the setting "All Active Threads" in Constant Throughput Timer. Then by using 7200/minute in Throughput timer you will not be able to maintain a load of 1000 users on your application, because when you are using Throughput Timer JMeter pauses threads in a way that you will get Target Throughput, which means using 7200/min (120/sec) with only 1 sampler, JMeter will generate load of 120 requests and not 1000 requests (which you want). So, either you have to increase the value in Timer or remove this timer, so that you are able to generate and put load of 1000 users on application.

Now, the question about the sessions remain active or not for the script duration (i.e. 3 minutes), depends upon both your application and server, i.e. the session timeout for application and server should be more than 3 minutes and no hold queue should be there on your server which holds the upcoming user requests till the previous requests are not complete, this hold queue should be more than 1000 + 100 = 1100 requests. You should talk to developer or the server configuration team for such configurations, they will be able to provide you correct idea about these settings and then you can plan your performance tests accordingly.