Ensure TaintBasedEviction int test not rely on TaintNodeByConditions #84036
Conversation
k8s-ci-robot
added
release-note-none
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: Huang-Wei The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
k8s-ci-robot
added
the
approved
| @@ -381,7 +381,8 @@ func nodeTainted(cs clientset.Interface, nodeName string, taints []v1.Taint) wai | |||
| return false, err | |||
| } | |||
|
|
|||
| if len(taints) != len(node.Spec.Taints) { | |||
There was a problem hiding this comment.
In terms of the testcase, if both TaintBasedEviction and TaintNodeByCondtions are enabled, there should be at least 2 taints (NotReady:NoExecute and NotReady:NoSchedule) applied, so comparing the size is absolutely incorrect.
And also the comment said "contain", the old logic doesn't follow either.
|
/retest |
k8s-ci-robot
added
the
lgtm
|
Thanks for the update, this test has been bugging me |
|
@Huang-Wei do you know if this change will fix #83321? |
|
@damemi Probably. The old testing logic actually was based on an enabled TaintNodeByCondition flag, while composing the node lifecycle manager by disabling it - which is inconsistent, so might cause issues. Let see how the test goes, it it still flakes, the most suspicious things could be the node lifecycle controller resync/gracePeriod settings, or the admission settings. |
|
@Huang-Wei thanks, I did notice before that TaintNodeByCondition was disabled, and reading through the code assumed this was intentional (as the exact taint we were expecting is added by TaintBasedEvictions) |
What type of PR is this?
A followup on #82703 (comment)
What this PR does / why we need it:
Right now, the TaintBasedEviction integration test relies on setting TaintNodeByConditions to false, which is not a reasonable assumption.
Which issue(s) this PR fixes:
Prerequisite of #82703.
Special notes for your reviewer:
Does this PR introduce a user-facing change?:
/sig scheduling
/cc @draveness @ahg-g
/kind bug