When we use the (current implementation) start time of the test plan, OMTG will think it is 10 seconds late in the game (that is how long the setup thread group made us wait) and start all samples concurrently.
With System#currentTimeMillis() as start time, we get the correct start time and the samples will be issued correctly. That is, they should be roughly issued with 1 req/s.
Expected behavior
Open model thread group should work the same when used with or without a long running setup thread group before it.
Actual behavior
When using open model thread group together with a long running setup thread group, we currently might generate sample storms.
Steps to reproduce the problem
Use the test plan contained in omtg-and-setup-group.jmx.zip
When we use the (current implementation) start time of the test plan, OMTG will think it is 10 seconds late in the game (that is how long the setup thread group made us wait) and start all samples concurrently.
With System#currentTimeMillis() as start time, we get the correct start time and the samples will be issued correctly. That is, they should be roughly issued with 1 req/s.
JMeter Version
5.7.3
Java Version
openjdk version "17.0.12" 2024-07-16
OS Version
Ubuntu 24.04.1