Repository navigation
Flaky Linux tests no longer flaky? #4446
Description
Activity
- addedquestionIssues asking questions about Node.js.Issues asking questions about Node.js.testIssues and PRs related to Node.js core tests and test infrastructure.Issues and PRs related to Node.js core tests and test infrastructure.
on Dec 28, 2015 Could you stress test them?
I couldn't find out which Linux nodes these tests were supposedly flaky on, so I'm just picking a few.
First up:
test-http-client-timeout-event:centos5-32: https://ci.nodejs.org/job/node-stress-single-test/nodes=centos5-32/239/ubuntu1504-64: https://ci.nodejs.org/job/node-stress-single-test/nodes=ubuntu1504-64/237/armv7-wheezy: https://ci.nodejs.org/job/node-stress-single-test/nodes=armv7-wheezy/240pi1-raspbian-wheezy(I'll never do that again until Cross-compile for (ARM) stress tests build#277 is resolved): https://ci.nodejs.org/job/node-stress-single-test/nodes=pi1-raspbian-wheezy/238
test-http-client-timeout-eventwas marked flaky because of failures on pi1-raspbian-wheezy. #2555test-child-process-bufferingwas marked flaky because of failures on armv7-wheezy.
#2627 (If it's not flaky anymore, that issue can be closed, of course.)Not sure what OS
test-http-client-timeout-eventtest-child-process-exit-codewas failing on so the scatter-shot method might need to be used there too. Although I don't think I've seen a yellow build in a long, long time on Linux, so it's probably not flaky now....@Trott With regard to your last remark, did you mean
test-child-process-exit-codeinstead oftest-http-client-timeout-event?@mscdex Yes, that's exactly right. It was a cut and paste error. Sorry about that. I'll edit it...
Ok, here's a stress test for
armv7-wheezyandtest-child-process-buffering: https://ci.nodejs.org/job/node-stress-single-test/247/nodes=armv7-wheezy/Here's a stress test for
pi1-raspbian-wheezyandtest-http-client-timeout-event(I accidentally aborted the one I had going for 3 hours... argh): https://ci.nodejs.org/job/node-stress-single-test/248/nodes=pi1-raspbian-wheezy/Looking good so far!
I've been looking back through the test history for the Linux CI nodes and I cannot find those tests ever failing in recent months.
I'd be good with removing these three tests at this point. The only one that hasn't been run through the stress test is
test-child-process-exit-code. Like you, I can't find any examples of it failing recently. If we remove it from the flaky list and it turns out that it is in fact still flaky (which seems unlikely to happen), we can put it back in as flaky.@Trott Sounds good to me.
- added a commit that references this issue
on Jan 3, 2016 - added a commit that references this issue
on Apr 2, 2016 - added a commit that references this issue
on Jul 27, 2026
Currently there are 3 tests marked as flaky on Linux:
They were all added in the initial flaky test list back in August with #2424. I've been looking back through the test history for the Linux CI nodes and I cannot find those tests ever failing in recent months. Perhaps they can be removed from the list now?