Stop deleting underlying services when federation service is deleted #37353
Conversation
k8s-ci-robot
added
the
cncf-cla: yes
nikhiljindal
added
area/cluster-federation
kind/bug
nikhiljindal
added
the
release-note
|
cc @saad-ali as FYI for 1.5. This is a bug fix for 1.5. Required in 1.5 since we are introducing cascading deletion based on DeleteOptions.OrphanDependents for federation resource. Federation service controller was doing it even without that option which is unexpected. |
k8s-github-robot
added
the
size/M
|
Brought up a cluster and verified that services are not deleted in underlying clusters and DNS records are not touched. |
| // or we do nothing for service deletion | ||
| // TODO: Should uncomment this? | ||
| // We should delete the load balancer when service is deleted | ||
| // Look at the same method in kube service controller. | ||
| //err := s.dns.balancer.EnsureLoadBalancerDeleted(service) |
There was a problem hiding this comment.
Remove this as well. We are not going to do this.
There was a problem hiding this comment.
umm. Why not? kube service controller calls this.
Leaving this as is. Can remove it in a separate PR if we want to.
There was a problem hiding this comment.
@nikhiljindal by "this" I meant the TODO comment and commented code. What was your "this"?
What loadbalancer do you think we are going to delete here? Also, the commented code is wrong anyway. s.dns is an interface that doesn't contain a field called balancer. My gut feeling is that, this was a copy-paste from cluster-level service controller and doesn't apply here.
Please remove it. The comment change is making things worse. And the right fix for that is just removing it.
There was a problem hiding this comment.
ah you are right. Removed it. Thanks for clarifying
| err := s.deleteClusterService(clusterName, cachedService, cluster.clientset) | ||
| if err != nil { | ||
| hasErr = true | ||
| } else if err := s.ensureDnsRecords(clusterName, cachedService); err != nil { |
There was a problem hiding this comment.
We should ensure that this behavior is carried over to the cascading deletion. Could you please add a TODO somewhere? Or open an issue may be? One of the two, either one is fine.
There was a problem hiding this comment.
I don't see the effect of s.ensureDnsRecords() in PR #36390. May be I am missing something?
| for clusterName, clusterClientset := range clusters { | ||
| _, err := clusterClientset.Core().Services(service.Namespace).Get(service.Name) | ||
| if err != nil { | ||
| framework.Failf("Unexpected error in fetching service %s in cluster %s, %s", service.Name, clusterName, err) |
There was a problem hiding this comment.
I am not entirely sure about failing here. It should log an error and continue?
Also, I would just put this in a defer block.
There was a problem hiding this comment.
aah the comment "Cleanup" was wrong. This is the test. We want to verify that services are not deleted from underlying clusters. Removed the comment.
AfterEach deletes these services.
nikhiljindal
force-pushed
the
serviceDelete
branch
from
82b5284
to
e2540b7
Compare
|
Thanks @madhusudancs Adding LGTM label |
nikhiljindal
added
the
lgtm
|
@nikhiljindal please look at the comment. |
nikhiljindal
force-pushed
the
serviceDelete
branch
from
e2540b7
to
34eae22
Compare
nikhiljindal
removed
the
lgtm
|
Thanks @madhusudancs |
nikhiljindal
added
the
lgtm
|
@k8s-bot test this |
|
@k8s-bot cvm gce e2e test this issue: #IGNORE (stuck on Waiting for status to be reported) |
|
@k8s-bot cvm gce e2e test this issue: #IGNORE (failed with "Error starting build") |
|
Removing requires release czar attention as per offline discussion with saad last week. |
|
Jenkins GCE e2e failed for commit 34eae22. Full PR test history. The magic incantation to run this job again is |
|
k8s-bot cvm gce e2e test this (build failed) |
|
@k8s-bot cvm gce e2e test this (build failed) |
|
@k8s-bot cvm gce e2e test this |
|
@k8s-bot gci gke e2e test this |
Ack. Thanks |
|
Unit tests seem to have been stuck for more than 11 hours. |
|
@k8s-bot test this [submit-queue is verifying that this PR is safe to merge] |
|
Automatic merge from submit-queue |
saad-ali
added
the
cherry-pick-approved
|
Commit found in the "release-1.5" branch appears to be this PR. Removing the "cherrypick-candidate" label. If this is an error find help to get your PR picked. |
Fixes #36799
Fixing federation service controller to not delete services from underlying clusters when federated service is deleted.
None of the federation controller should do this unless explicitly asked by the user using DeleteOptions. This is the only federation controller that does that.
cc @kubernetes/sig-cluster-federation @madhusudancs
This change is