Am 07.04.24 um 09:42 schrieb jean-frederic clere:
On 4/6/24 20:02, Rainer Jung wrote:
Hi Jean-Frederic and all,
you didn't write at what point in time you take the thread dump. I see
the SIGTERM messages logged during test execution always during the
last test in each group (http2, md, ...) just because that is the time
the logs are checked by teardown for error messages. At the time the
test complains it already starts to kill the children and at least
during my test runs it success with killing them (I think). So finding
a good point in time to attach the debugger and see the right
situation might not be easy?
I might have taken the thread dump late because I have taken it just
after the ap_log_error("sending a SIGTERM") at that time there is only
one thread running and it is "waiting" for the listener thread that has
already stopped. There I have found that the behavior of pthread_kill()
changed in "recent" fedora/rhel.
Are you suggesting I should also try to get dumps earlier in the
shutdown process?
When you say Yann's patch helps, it means especially there are not
more SIGTERM messages in the logs resp. no more teardown checks failing?
Yes with Yann's patch, the "AH00045: child process nnnnn still did not
exit, sending a SIGTERM" messages are gone and teardown checks are passing.
I could now also do some tests and for me the SIGTERM messages for
worker on RHEL 9 during pytest are also gone with Yann's woker MPM
patch. Plus I didn't get any new failures. Perl test framework is fine
and pytest only some of the few occasional failures and errors I had
also seen before.
So it would be nice if we would apply Yann's patch for the worker mpm.
Thanks and best regards,
Rainer
