https://bz.apache.org/bugzilla/show_bug.cgi?id=64618
Philippe Mouawad <[email protected]> changed: What |Removed |Added ---------------------------------------------------------------------------- Hardware|PC |All Resolution|--- |WONTFIX Target Milestone|JMETER_5.4 |--- Status|REOPENED |RESOLVED --- Comment #3 from Philippe Mouawad <[email protected]> --- (In reply to tdbgauthamkumar from comment #2) > This bug is happening in JMeter 5.3 version. > Same parallel controller scripts working when downgraded to JMeter 5.2.1 > version. > > So, Please fix it as its been a blocker/critical bug for our jmeter > performance pipeline where we distribute to simulate millions of users. This plugin has IMO, a very hacky approach to do its work: - Using reflection to access internal jmeter core fields https://github.com/Blazemeter/jmeter-bzm-plugins/blob/master/parallel/src/main/java/com/blazemeter/jmeter/controller/JMeterThreadParallel.java#L44 - Making Non thread safe core JMeter components "thread safe" by cloning them https://github.com/Blazemeter/jmeter-bzm-plugins/blob/master/parallel/src/main/java/com/blazemeter/jmeter/controller/ParallelSampler.java#L338 ... What is happening here is that the Samplers are sharing HTTPHC4Impl#HTTPCLIENTS_CACHE_PER_THREAD_AND_HTTPCLIENTKEY and randomly, one thread triggered by this plugin will close HTTPClient used by another one leading to the problem. I would suggest you report this issue to the maintainer of this plugin. It's not because it works with JMeter 5.2 and not in 5.3 that issue comes from 5.3. It's very possible that approach of plugin relied on a hack that does not work anymore. Anyway, JMeter team is composed of volunteers working on their personal time and sometimes with some "sponsoring". We spend most of our time on Core and it's quite enough. As you run millions of users, I guess you can afford paying somebody to dig in this issue to see where it comes from and propose a patch. -- You are receiving this mail because: You are the assignee for the bug.
