-
Notifications
You must be signed in to change notification settings - Fork 40k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
MESOS: Refactor scheduler #16316
MESOS: Refactor scheduler #16316
Conversation
Labelling this PR as size/XXL |
GCE e2e build/test failed for commit 6774881. |
6774881
to
a2e34a3
Compare
GCE e2e test build/test passed for commit a2e34a3. |
GCE e2e test build/test passed for commit 629667141a9ca1307add6d9be6e39d1274756332. |
6296671
to
4d3c6ce
Compare
GCE e2e test build/test passed for commit 4d3c6cea59e451f506e599a388377ea8efc7722a. |
78c4a7f
to
3a292f0
Compare
f3e9938
to
cdd94dc
Compare
/cc @spacejam |
GCE e2e test build/test passed for commit 78c4a7fca2b54e0a85aa7a82d8a1b55ea19d0ab3. |
GCE e2e test build/test passed for commit 3a292f0442e52aabfa41d736c1af1b782c58fd25. |
GCE e2e test build/test passed for commit f3e9938bce04d89139d998f02c0cab2e69f5218f. |
GCE e2e test build/test passed for commit cdd94dc8579c7d8be73b602a107369b7e960384c. |
GCE e2e test build/test passed for commit 9fc47619de829e5807df1b3e0eca48b8229f9092. |
@@ -832,131 +806,3 @@ func TestPlugin_LifeCycle(t *testing.T) { | |||
time.Sleep(time.Second / 2) | |||
failPodFromExecutor(launchedTask.taskInfo) | |||
} | |||
|
|||
func TestDeleteOne_NonexistentPod(t *testing.T) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Has this test became untestable due to visibility or was it moved too?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Moved into deleter_test.go
GCE e2e test build/test passed for commit e71f43d. |
Submit queue seems to be blocked with this one for days, showing "Github CI tests are not green." which is false. Also the cla/google test is not triggered. Something is pretty broken here. |
Merging manually as this is completely contrib/mesos related and submit queue ignores it. |
This needed a squash. |
We decided against squashing intentionally. A lot of those commits are renamings and movements of code. After squashing it would be nearly impossible to trace code into the code base before the refactoring. I hope that's ok in this special situation. In other cases, I completely agree. |
@mikedanese another question, I think you mentioned something some time ago: why was this PR always ignored by the merge bot, even with the e2e-not-required label? Is this always the case with contrib-only PRs? It showed "Github CI tests are not green." although everything was green. Any idea? |
If you go to http://submit-queue.k8s.io/ and search your username or PR number in the search box it should tell you what merge bot thinks the current status of the pr is |
That's what I mean. It was staying in "Github CI tests are not green." although everything was green. |
A commit message would have been a good place to trace renaming. This also broke unit tests. Squash commits in the future and allow mergebot to merge PRs. |
@mikedanese sorry for the trouble. To avoid this next time I really would like to understand: It's not that we merged here immediately. We were waiting for this PR two days, fulfilling all criteria for a mergeable PR. The submit queue didn't pick it up or changed it's status. In fact we had the same situation before with a contrib-only PR and I think you or gmarek clicked the merge button eventually. |
I think it may have been rightfully holding off on merging because of unit test failures since submit-queue.k8s.io was giving a correct reading, but not sure why it wasn't surfaced in the PR status. Right now you should look for jenkins/k8s-e2e, jenkins/k8s-unit, and travis-ci to be green. @ixdy if the unit test status isn't showing up on a PR is there a way to kick jenkins? Same way as e2e tests? If you think that k8s-merge-bot is not working, file an issue in contrib/, ping @eparis or @lavalamp on the PR or in k8s-dev slack. |
@sttts Next time, please ask @k8s-oncall to merge things instead of merging them yourself--you don't know what the build cop might be in the middle of doing, an unexpected merge can make their life difficult (imagine they're in the process of fixing the build). Plus if merge queue is ignoring something it shouldn't, we'd like to catch it in the act. I will also complain about the excessive number of commits here. At a minimum, things like "fix rebase error", "gofmt fixes", "fix comment", etc., should be squashed with the commits that introduced the error(s). I buy keeping separate commits to ease rebasing when doing a major rename, I have done that myself. I buy keeping separate commits when there's a possibility you might want to roll them back independently. I buy commits like in this PR to make review easier, but I think they should be squashed before merge. Please see:
I think this is a sausage and should have been squashed at least partially, maybe if it were more like ~10 commits I wouldn't be writing this. Merging it this way leaves us a lot of extra stuff to wade through when e.g. making release notes. I realize this is a bit un-git-ish of us, but it's what we've gotten consensus on. |
@lavalamp thanks for the hint about @k8s-oncall. This will help in those situations. About squashing, I totally get the argument about "wading through" for release notes and the gotten consensus in order to keep the repo manageable. We will keep that in mind next time we do something similar. |
@mikedanese saying "@k8s-bot unit test this" will trigger unit tests only. |
additionally, the "test this" phase actually triggers both e2e and unit tests now. |
Still thinking unit tests were running: "Jenkins unit/integration 2381 tests run, 8 skipped, 0 failed.". I guess that's the jenkins/k8s-unit job in Jenkins. Unfortunately, I cannot access the logs from the link above. Can anybody confirm that the contrib unit tests are run in that job? |
It looks like unit tests ran and passed, yeah. I think you raced with #16992, which got merged at approximately the same time; both individually passed unit tests, but after merging there was a build break. |
If you like reading XML or overly-verbose output, you can see which tests were run at https://console.developers.google.com/storage/browser/kubernetes-jenkins/pr-logs/e71f43de930c15c78c71c454836424ddf2829d42/kubernetes-pull-test-unit-integration/1450/artifacts/. The build log is also supposed to be uploaded, but for some reason it was not (which I'm working on fixing). The submit queue re-runs all tests before merging, so it likely would have caught this merge race. |
@ixdy thanks for the investigation and for restoring my world view. The submit queue just got a big new fan. The contrib unit tests were actually running. The theory with the race makes sense. |
MESOS: Refactor scheduler
MESOS: Refactor scheduler
MESOS: Refactor scheduler
MESOS: Refactor scheduler
Main goals
Steps: