Fast scale up when an asg fault

[daimaxiaxie]

What type of PR is this?

/kind bug

What this PR does / why we need it:

when an asg fault(such as out of stock), Cluster Autoscaler should recovery quickly. Then scale up other asg.

Which issue(s) this PR fixes:

Fixes #6645, #6128

Special notes for your reviewer:

If the instance has error, add it to LongUnregistered immediately and then delete instance. Merge deleteCreatedNodesWithErrors into removeOldUnregisteredNodes.

Does this PR introduce a user-facing change?

None

Additional documentation e.g., KEPs (Kubernetes Enhancement Proposals), usage docs, etc.:

[k8s-ci-robot]

Hi @daimaxiaxie. Thanks for your PR.

I'm waiting for a kubernetes member to verify that this patch is reasonable to test. If it is, they should reply with /ok-to-test on its own line. Until that is done, I will not automatically test new commits in this PR, but the usual testing commands by org members will still work. Regular contributors should join the org to skip this step.

Once the patch is verified, the new status will be reflected by the ok-to-test label.

I understand the commands that are listed here.

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.

[towca]

/ok-to-test

[MaciekPytel]

This does a pretty significant refactor of error handling mechanism, that doesn't seem strictly necessary for fast fail path in AWS. At the same time it's pretty risky - the way AWS provider does fake nodes there may be no difference between deleteCreatedNodesWithErrors and removeOldUnregisteredNodes, but how fake nodes behind the scenes vary widely between providers.
If we want to make changes like that, I'd like a strong and clear argument that this will not make a difference that is observable by any provider, regardless of how it implements fake nodes behind the scenes (e.g. in GCE there are no fake nodes at all, GCE MIG will really have instance objects in error states that must be cleaned up by CA).

[daimaxiaxie]

Makes sense, I made some changes. @MaciekPytel

deleteCreatedNodesWithErrors only delete State == cloudprovider.InstanceCreating && ErrorInfo != nil node.
Now removeOldUnregisteredNodes only delete State = cloudprovider.InstanceCreating && ErrorInfo != nil node. Therefore, all cases of deleteCreatedNodesWithErrors can be covered(except MinSize, but normal scale up).

And State = cloudprovider.InstanceCreating && ErrorInfo != nil node join LongUnregistered has no effect on any provider.

This mechanism is not only valid for AWS, but also for other providers, such as Azure. Does not cause a difference that is observable by any provider, because it just moves the existing logic. This is also healthier than now.

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by: daimaxiaxie
Once this PR has been reviewed and has the lgtm label, please assign feiskyer for approval. For more information see the Kubernetes Code Review Process.

The full list of commands accepted by this bot can be found here.

Needs approval from an approver in each of these files:

• cluster-autoscaler/OWNERS

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[k8s-ci-robot]

PR needs rebase.

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-sigs/prow repository.