I've done a lot of tests in JMeter for REST Api using HTTP Requests and sql testing with JDBC Requests. Now the problem is that in every test first request gets even 10x higher response time than others. I know that this is due to JMeter need to establish connection to server or database. When i use more threads for example 200 this problem exists for first requests in every threads.
Is there any good solution in JMeter to ignore that connect time in computing average/deviation/throughput?
3
votes
What version of JMeter are you using and for how much time are your running your test ?
- UBIK LOAD PACK
@UBIKLOADPACK I run 12,000 tests and for example for first shot i get 7000ms and for second 50ms so its actually even more than 100x. Maybe its not big deal for computing average but i think for deviation it matters. For now i am thinking about running tets with for example 1000 starting request which should be enough time establish all connections and remove those first 1000 results before computing average and deviation, what do you think about it, have you better idea? I use Apache 3.0.
- jgr
@jgr Did you find any solution for this issue?I'm facing same issue in my DB test,First request always take 10-15 sec more then actual time.
- Devang
@TestAuto check my answer below maybe it will help you
- jgr
1 Answers
0
votes
I finally ended up making some introductory requests in each of my test plan which i did not take into account in final results.
For example i had 12000 requests in test plan for queries phrases from csv file. But first 2000 phrases was for example for phrase 'test'. That 2000 made me sure that all threads will be up. My results were saved in csv file so after tests it was easy to cut off them, then read file in JMeter and compute statistics without those 2000.
