AWSMachinePool does not drain nodes during scale-in

[dthorsen]

/kind bug

What steps did you take and what happened:

• Create a workload cluster with the experimental EKS Control Plane.
• Create a MachinePool with replicas: 5 and create the associated AWSMachinePool resources. (Note: this AWSMachinePool is not managed by cluster-autoscaler)
• Create a deployment and scale it such that some pods fall on all machines
• Create a PDB protecting the deployment with maxUnavailable: 1
• Scale the MachinePool down to replicas: 3

This caused the AWSMachineController to set the DesiredInstances in the ASG to 3 without draining nodes at all. The PDB was not honored, and the EC2 instances were terminated by the ASG immediately.

What did you expect to happen:
The nodes should have drained gracefully before the EC2 instances are terminated.

Anything else you would like to add:
In the current AWSMachinePool implementation, the instance selection for scale-in is performed at the AutoScalingGroup. This could be fixed in the non-cluster-autoscaler case by modifying AWSMachinePool controller to perform node selection for scale-in, drain the selected nodes, and finally utilize the AWS TerminateInstanceInAutoScalingGroup action while setting the request value ShouldDecrementDesiredCapacity: true

We may want to also consider a lifecycle hook on the autoscaling group that prevents ec2 instance termination until the drain completes. This would help to prevent cases where instances are forcibly terminated without draining when the DesiredInstances values are manipulated via the EC2 console, CLI, or APIs.

Environment:

Cluster-api-provider-aws version: Commit: 3338cd4
Kubernetes version: (use kubectl version): v.1.17.9
OS (e.g. from /etc/os-release): Amazon Linux 2

[fejta-bot]

Issues go stale after 90d of inactivity.
Mark the issue as fresh with /remove-lifecycle stale.
Stale issues rot after an additional 30d of inactivity and eventually close.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle stale

[fejta-bot]

Stale issues rot after 30d of inactivity.
Mark the issue as fresh with /remove-lifecycle rotten.
Rotten issues close after an additional 30d of inactivity.

If this issue is safe to close now please do so with /close.

Send feedback to sig-contributor-experience at kubernetes/community.
/lifecycle rotten

[randomvariable]

Chatting with @sedefsavas
AWS Node Termination Handler ( https://github.com/aws/aws-node-termination-handler ) can help, but doesn't fully eliminate it - it gives a 2 minute warning.

Sync with CAPZ on MachinePool v.Next

@kschumy , any ideas on what we should do here?

[sedefsavas]

We can follow a similar approach with Openshift's POC about polling termination endpoint:
https://github.com/openshift/cluster-api-provider-aws/blob/b4a3478db44ddb554883cf77a9e5f49ffd54fdf4/pkg/termination/handler.go

More on this is discussed in the cluster-api proposal: kubernetes-sigs/cluster-api#3528

[fejta-bot]

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-contributor-experience at kubernetes/community.
/close

[k8s-ci-robot]

@fejta-bot: Closing this issue.

In response to this:

Rotten issues close after 30d of inactivity.
Reopen the issue with /reopen.
Mark the issue as fresh with /remove-lifecycle rotten.

Send feedback to sig-contributor-experience at kubernetes/community.
/close

Instructions 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/test-infra repository.

[richardcase]

/reopen
/remove-lifecycle rotten

[richardcase]

From office hours 2023-04-03:

• This will potentially be handled by Implement MachinePool Machines clusterAPI proposal #4184.
• Providers refresh will have a weakness as aws only give a small amount of time before termination (same issue with AWSManagedMachinePools)
• Users expectation is that nodes are drained

/triage accepted
/priority important-soon

[dlipovetsky]

Also from office hours discussion:

Users define Pod Disruption Budgets to ensure that their Pods are not voluntarily deleted.

A scale-in of a MachinePool, if it uses the "providers refresh", will always proceed, even if it violates a budget.

For comparison, a scale-in of a MachineDeployment will never proceed if it violates a budget.

[k8s-triage-robot]

This issue is labeled with priority/important-soon but has not been updated in over 90 days, and should be re-triaged.
Important-soon issues must be staffed and worked on either currently, or very soon, ideally in time for the next release.

You can:

• Confirm that this issue is still relevant with /triage accepted (org members only)
• Deprioritize it with /priority important-longterm or /priority backlog
• Close this issue with /close

For more details on the triage process, see https://www.kubernetes.dev/docs/guide/issue-triage/

/remove-triage accepted

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle stale
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[k8s-triage-robot]

The Kubernetes project currently lacks enough active contributors to adequately respond to all issues.

This bot triages un-triaged issues according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue as fresh with /remove-lifecycle rotten
• Close this issue with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle rotten