devinbost commented on issue #9916: URL: https://github.com/apache/pulsar/issues/9916#issuecomment-814581870
I just got this test in #10154 The fundamental problem with this test (and many of the other Flaky tests) is that we are basing test logic on timing, which falls apart when the test runner non-deterministically encounters a "stop the world" garbage collection event (or some other kind of event that creates random delays) somewhere in the middle of that. If the test runner is very resource-constrained, those kinds of events will happen a lot as the JVM tries to free up resources. A large testbed that has lots of temporary one-time-use objects like Pulsar's is the perfect breeding ground for generating lots of dead objects that need to be garbage collected, resulting in long garbage collection times, further exasperating the issue. Sharing resources between tests may help with that, but it's not a perfect solution because it would be impossible to completely eliminate garbage collection in the test runner. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
