Skip to content

Cordon nodes during scale down by default #8176

@x13n

Description

@x13n
Member

Which component are you using?:

/area cluster-autoscaler

Is your feature request designed to solve a problem? If so describe the problem this feature should solve.:

Cluster Autoscaler has a --cordon-node-before-terminating flag, but it is set to false by default. I think we should always cordon nodes before deleting them.

Describe the solution you'd like.:

  • (short term) Flip the default flag value to true. Mark the flag as deprecated & planned for removal.
  • (in 1-2 minor releases) Delete the flag.

Describe any alternative solutions you've considered.:

Keep as is. Or keep the flag. I didn't find any good reason to do any of these things.

Additional context.:

Lack of cordoning has been a pain point since forever (#1136, #2505, #2868, #3648). It was later added as an option hidden behind a flag, but it could just work out of the box instead of requiring manual setup.

Activity

added
kind/featureCategorizes issue or PR as related to a new feature.
on May 28, 2025
x13n

x13n commented on Jun 2, 2025

@x13n
MemberAuthor

/hold

Putting on hold so it lands in 1.34 release, not 1.33, as discussed on the SIG meeting.

x13n

x13n commented on Jun 2, 2025

@x13n
MemberAuthor

Whoops, meant to add this on the PR, not the issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

      Development

      Participants

      @x13n@k8s-ci-robot

      Issue actions

        Cordon nodes during scale down by default · Issue #8176 · kubernetes/autoscaler