On Sat, 20 Dec 2025 at 10:18, David Bolen <[email protected]> wrote: > > While unlikely to be the issue, there was a problem with test_profiling > aborting on my buildbot back from Dec 8-11, due to an interaction with an > earlier commit and auditing (fixed by Victor in > https://github.com/python/cpython/issues/142627). > > What was happening is that when test_profiling aborted, it stranded one or > two processes that would hang around forever and eat up CPU. The most common > one was a process (owned by init) running a command line python -c script > (involving ExpensiveGarbage()) that seemed stuck with their sockets in > CLOSE_WAIT presumably in some sort of tight loop burning CPU. Over time they > ate up some other sort of resource as every build would start failing with > odd errors. > > In that case, the issue was not directly related to test_profiling, which may > or may not be the case here, but I do think something in that test suite is > not friendly if the process aborts mid-test without cleanup. It was a > SIGABRT in my case, and looks like SIGSEGV here in your build, but probably > the same result. > > While debugging the issue I did in fact implement a monitor script to kill > off any such processes, but I haven't seen anything show up on my buildbot > since the 12th. And it looks like test_profiling was fine on my equivalent > build matching your 12988, so not quite sure what's to make of that. >
Cool! Thanks for the additional info. I'll wait and watch for the time being; if it happens again, I'll follow up here and look into making a process killer. Probably if a process is consuming 100% CPU for an hour, it can be killed. ChrisA _______________________________________________ Python-Buildbots mailing list -- [email protected] To unsubscribe send an email to [email protected] https://mail.python.org/mailman3//lists/python-buildbots.python.org Member address: [email protected]
