You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Investigate flaky crash in test/pummel/test-repl-empty-maybelocal-crash.js on freebsd #42719
@targos if you're talking about the freebsd error in this CI run - #41513 (comment), i.e., the java.nio.channels.ClosedChannelException error, I think that's different from this one. In this issue, it's just the node process that crashes, and the runner still remains connected to the CI server. (FWIW, we faced that error on arm for this PR, discussed here - #42409 (comment))
I have suspicions that the new test is causing the disconnects on fedora-latest, e.g. https://ci.nodejs.org/job/node-test-commit-linux/nodes=fedora-latest-x64/45410/console -- it looks like the OOM Killer is stepping in. I don't have definitive proof it is this test, but we're in the area of executing the pummel tests and this has only started happening recently.
Also this is different from the NIO disconnects we've seen elsewhere caused by Jenkins ping timeouts -- there's no corresponding ping thread timeout message in the Jenkins server log.
Activity
targos commentedon Apr 13, 2022
The issue seems similar to #41513 (review)
RaisinTen commentedon Apr 13, 2022
@targos if you're talking about the freebsd error in this CI run - #41513 (comment), i.e., the
java.nio.channels.ClosedChannelExceptionerror, I think that's different from this one. In this issue, it's just thenodeprocess that crashes, and the runner still remains connected to the CI server. (FWIW, we faced that error on arm for this PR, discussed here - #42409 (comment))richardlau commentedon Apr 13, 2022
I have suspicions that the new test is causing the disconnects on fedora-latest, e.g. https://ci.nodejs.org/job/node-test-commit-linux/nodes=fedora-latest-x64/45410/console -- it looks like the OOM Killer is stepping in. I don't have definitive proof it is this test, but we're in the area of executing the pummel tests and this has only started happening recently.
journalctl logs
The droplet has 4 GB of memory.
Also this is different from the NIO disconnects we've seen elsewhere caused by Jenkins ping timeouts -- there's no corresponding ping thread timeout message in the Jenkins server log.
richardlau commentedon Apr 14, 2022
Also crashed on RHEL 8 x64:
test: delete test/pummel/test-repl-empty-maybelocal-crash.js
BethGriggs commentedon Apr 14, 2022
Landed in 19064be
test: delete test/pummel/test-repl-empty-maybelocal-crash.js
test: delete test/pummel/test-repl-empty-maybelocal-crash.js
test: delete test/pummel/test-repl-empty-maybelocal-crash.js
test: delete test/pummel/test-repl-empty-maybelocal-crash.js
test: delete test/pummel/test-repl-empty-maybelocal-crash.js