Hello, My vote: [-1] for the explained reason Thanks for RM and all the work from Vladimir and Felix. My answers inline below.
Regards On Thu, Jun 22, 2023 at 10:59 PM Vladimir Sitnikov < sitnikov.vladi...@gmail.com> wrote: > > I think there is a behavioural change related to this commit: > > > it seems to me we can end up losing > SampleResults > > We are stopping the test, and I doubt we can guarantee we will record all > events, including those that happen exactly at the moment of stopping the > test. > > The test case I added shows that "interrupted" event is generated. If > BackendListener loses events, it might be an issue in BackendListener code. > Since queue.offer is interrupted, the SampleResult is lost > > --- > > The change was a bug fix. > Before the fix it was not possible to terminate (e.g. stop) a test if http > sampler was waiting the response from server (e.g. if server was stuck) > This issue was always there due to missing timeout on HTTP Sampler no ? > > Note, that OMTG was already using java thread interrupt previously, and now > the key change I made was **jmeter** thread interrupt (jmeterThread has its > own interrupt method). > For me previously we Safe stopped the thread without interrupt. Calling Interrupt introduces a regression in BackendListener which is a component that is used a lot I think. > It would be great if you could file an issue that describes the regression, > however, if people need to wait some time at the end of the test, they can > use `pause(2 min)`, so OMTG will wait for 2 min before interrupting > samplers. > I'll do that > > We probably need something like `await_completion(5 min)` which would wait > till all threads to complete 5 min maximum, and then interrupt them. > > > Vladimir > > -- > Vladimir > -- Cordialement. Philippe Mouawad.