-
Notifications
You must be signed in to change notification settings - Fork 40.6k
Add sleep to EtcdTestServer Terminate. #19187
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
Conversation
Labelling this PR as size/XS |
GCE e2e test build/test passed for commit 4e08c8ad64b6fef094cf3ee700cef1bb2b7c7d63. |
@k8s-bot unit test this please |
8 similar comments
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
4e08c8a
to
e41d66f
Compare
GCE e2e test build/test passed for commit e41d66f0e207f8d32268b80e8d0d1c227663cc48. |
@k8s-bot unit test this please |
4 similar comments
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
It seems that we need to rebuild the test image too... |
@k8s-bot unit test this please |
5 similar comments
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
@k8s-bot unit test this please |
GCE e2e build/test failed for commit c8ad311. |
@k8s-bot test this please |
GCE e2e test build/test passed for commit c8ad311. |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
My best guess is that our jenkins machines are loaded way more than any workstation we use (and I doubt anyone actually tried to run it when machine is heavily loaded). |
GCE e2e build/test failed for commit c8ad311. |
@timothysc My guess is that jenkins is slow. What we did is to cancel all pending requests and return to make sure there is no in-flight requests. However, this is not robust since the request might already have arrived the server side and it is not cancellable at the moment. When this happens, there is a data race. The possibility of this increases when the latency between request+response increases. To solve this @wojtek-t sleeps a few here to wait for server finishing all pending requests. We can do better by accounting pending TCP connection at client side and waits for all TCP level shutdown. But I am not sure if it worth the effort. We should not wait for TCP shutdown in reality anyway. |
@k8s-bot test this |
GCE e2e build/test failed for commit c8ad311. |
@k8s-bot test this |
GCE e2e test build/test passed for commit c8ad311. |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
We have no barrier synchronization on test-server shutdown so we're band-aiding with sleep calls. @gmarek could someone pastebin the stack calls. Or provide a link which I can access to view? |
They are in the issue mentioned in the first comment: #18928 |
GCE e2e test build/test passed for commit c8ad311. |
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
Merging to deflake tests. |
Add sleep to EtcdTestServer Terminate.
GCE e2e test build/test passed for commit c8ad311. |
Hack to help with #18928, we should write a proper fix ASAP. @xiang90