NEEDS-JIRA: Fix node.kubernetes.io/exclude-from-external-load-balancers on masters#127
NEEDS-JIRA: Fix node.kubernetes.io/exclude-from-external-load-balancers on masters#127mdbooth wants to merge 5 commits into
Conversation
Cherry-picked from upstream kubernetes/cloud-provider-gcp commit d3a674c
…akeGCECloud NewFakeGCECloud sets fields on the Cloud struct from TestClusterValues but was missing the mapping from vals.SubnetworkURL to gce.unsafeSubnetworkURL. This caused g.SubnetworkURL() to return an empty string in tests that set vals.SubnetworkURL, which in turn made subnetNameFromURL() fail and skip node filtering entirely in ensureInternalInstanceGroups. This is the minimum required change from upstream commit b8caeb9 (PR kubernetes#647: "Add syncDefaultPodRanges change for Multi-networking default Param and Network"), which introduced SubnetworkURL into TestClusterValues and wired it through NewFakeGCECloud. Only that wiring is extracted here; the rest of PR kubernetes#647 is unrelated to this branch.
…balancers on masters We previously used allHaveNodePrefix() to ensure that if a master instance group contained the bootstrap machine this would not cause it to be excluded from re-use. This worked because the bootstrap machine's name has the same prefix as all other machines in the cluster. However, this logic unconditionally always included the master instance group. If the masters are labelled with node.kubernetes.io/exclude-from-external-load-balancers the CCM framework code will have filtered the Nodes before passing them to provider-GCP. But this logic meant we added them anyway, sending unintended traffic to the control plane machines. Manual backport of 52387cc
Commit f867d9c (UPSTREAM: 306: Disable GCE external load balancer when services handled by Ingress-GCE) accidentally removed the v1 "k8s.io/api/core/v1" import from gce_loadbalancer_external_test.go while the file continued to use v1.* types throughout, causing a build failure. The upstream fix is in commit 2e52b88 (Merge pull request kubernetes#930 from upodroid/build-binaries-correctly), which rewrote the entire file. This commit extracts only the minimum change required: restoring the missing import.
openshift-ci
Bot
added
the
do-not-merge/hold
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Repository: openshift/coderabbit/.coderabbit.yaml Review profile: CHILL Plan: Enterprise Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
|
[APPROVALNOTIFIER] This PR is NOT APPROVED This pull-request has been approved by: The full list of commands accepted by this bot can be found here. DetailsNeeds approval from an approver in each of these files:Approvers can indicate their approval by writing |
|
@mdbooth: The following test failed, say
Full PR test history. Your PR dashboard. DetailsInstructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository. I understand the commands that are listed here. |
1 participant
We previously used allHaveNodePrefix() to ensure that if a master instance group contained the bootstrap machine this would not cause it to be excluded from re-use. This worked because the bootstrap machine's name has the same prefix as all other machines in the cluster.
However, this logic unconditionally always included the master instance group. If the masters are labelled with node.kubernetes.io/exclude-from-external-load-balancers the CCM framework code will have filtered the Nodes before passing them to provider-GCP. But this logic meant we added them anyway, sending unintended traffic to the control plane machines.
Manual backport of #121
Additionally backports some small test fixes required for the tests to compile and pass. Required so we can execute the new test added for the bugfix.
/hold until the parent PRs merge