Created attachment 34274 [details] jmeter log, TPS graph generated by the reporting dashboard When the granularity is set below 1 second, the TPS and the other throughput graphs will be incorrect. For example if the granularity is 500ms, the TPS will be doubled on the graph. Steps to reproduce: 1. Set jmeter.reportgenerator.overall_granularity = 500 2. Run some tests (the attached example runs a TCP Sampler based custom sampler for the EPP protocol) with Reporting dashboard generated. Here you can see in the jmeter log (attached) the average TPS: ... 2016/09/19 05:24:00 INFO - jmeter.reporters.Summariser: summary + 33000 in 00:00:26 = 1246.7/s Avg: 53 Min: 30 Max: 118 Err: 0 (0.00%) Active: 70 Started: 70 Finished: 0 2016/09/19 05:24:00 INFO - jmeter.reporters.Summariser: summary = 33003 in 00:00:37 = 890.0/s Avg: 53 Min: 30 Max: 118 Err: 0 (0.00%) 2016/09/19 05:24:30 INFO - jmeter.reporters.Summariser: summary + 37300 in 00:00:30 = 1244.5/s Avg: 55 Min: 30 Max: 200 Err: 0 (0.00%) Active: 70 Started: 70 Finished: 0 2016/09/19 05:24:30 INFO - jmeter.reporters.Summariser: summary = 70303 in 00:01:07 = 1048.5/s Avg: 54 Min: 30 Max: 200 Err: 0 (0.00%) 2016/09/19 05:25:00 INFO - jmeter.reporters.Summariser: summary + 37200 in 00:00:30 = 1237.2/s Avg: 55 Min: 30 Max: 165 Err: 0 (0.00%) Active: 70 Started: 70 Finished: 0 2016/09/19 05:25:00 INFO - jmeter.reporters.Summariser: summary = 107503 in 00:01:37 = 1106.9/s Avg: 54 Min: 30 Max: 200 Err: 0 (0.00%) ... 3. Wait for the tests to finish and check the Transaction Per Second graph in the generated Reporting Dashboard Expected results: The same result should be in the dashboard report, around 1200 TPS. Actual results: The actual result in the graph seems to be doubled, around 2400 TPS. Additional information: - The full generated dashboard or the test result was above the 1MB limit, but I can send if it would be helpful. Right now I attached the jmeter.log and the TPS graph image. - If I set the granularity to above or equal to 1 second, the graph will be correct, so this is what I am doing currently.
Hello, Could you provide an overview of your test plan structure ? Could you test with nighlty build which has many fixes on report/dashboard ? If possible could you provide a simple test plan using java request with sleep time so that we can easily work on issue ? Thanks
Hello, I investigated the problem. Computing is ok, the issue is that the /sec graphs don't expect granularity below 1sec. The doubling you see is due to the fact that computation is made for 0.5 sec , that's why you see it double. We have 2 ways to fix this: - Either adapt graph to change axis labels, title ... - Or forbid granularity to be under 1 second @Team, thoughts on that ? Thanks
I think for now we should reject value lower than 1s. Moving to enhancement
This issue has been migrated to GitHub: https://github.com/apache/jmeter/issues/4118