We have run a performance test using blazemeter with jmeter script. In summary load graph is showing inaccurate data some one can explain?

Question

We recently conducted performance testing on BlazeMeter with 25,000 virtual users (VUs) within a 5-minute time frame. Unfortunately, approximately 3,000 VUs failed during the test. Upon reviewing the summary page after the test, we observed a graph that showed a decrease to 14,000 VUs after 15 minutes, followed by a gradual increase to 22,000 VUs.

We would appreciate it if someone could explain the behavior we observed in BlazeMeter. While we have already contacted the BlazeMeter team for clarification, they have attributed the issue to the engines reaching 90% CPU usage. While we agree that this may have caused the initial failure of approximately 3,000 VUs, we are still uncertain as to why the number of VUs decreased to 14k VUs and again increased gradually.

Max Users:

Below is the reference image :

Answer 1

Unfortunately we cannot provide a comprehensive answer by looking at these 2 incomplete charts and seeing your test configuration.

There is Logs Report tab in BlazeMeter which contains execution logs (including engine logs), there you will be able to find jmeter.log file which should normally contain information regarding threads start and stop event including the reason for stopping.

High CPU usage on the engine side is definitely a problem as you might get false negative results but it shouldn't cause threads termination.