Add validation for instanceType and ami architecture

[bharath-123]

When an ami and instanceType of different architectures are specified,kops update cluster does not fail even though its an invalid combination.

This can cause side effects like creation of unwanted versions of a launch template, failed cluster creations and so on.

This PR adds checks to validate the chosen instance type architecture against the ami architecture. This PR adds the validation for launch configuration, launch templates and mixed instance policies.

Open question: We are making a DescribeInstanceType call for each instance when validating for mixed instance policies. This can be expensive as more instances are added. Should we feature flag this validation?

UPDATE: I think the number of calls to DescribeInstanceType should be fine as this done only during spec validation and not during cluster creation.

Fixes: #10726

[k8s-ci-robot]

Hi @bharath-123. 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.

[hakman]

/ok-to-test

[hakman]

Quick thought @bharath-123, when we debated validation of instance types, we decided to do it at config validation time:

kops/pkg/apis/kops/validation/aws.go

Lines 121 to 132 in a140168

	func awsValidateInstanceType(fieldPath *field.Path, instanceType string, cloud awsup.AWSCloud) field.ErrorList {
	allErrs := field.ErrorList{}
	if instanceType != "" && cloud != nil {
	for _, typ := range strings.Split(instanceType, ",") {
	if _, err := cloud.DescribeInstanceType(typ); err != nil {
	allErrs = append(allErrs, field.Invalid(fieldPath, typ, "machine type specified is invalid"))
	}
	}
	}
	return allErrs
	}

Probably we should do the same for images and reduce the number of API calls later, when applying changes. I think the instance types are even cached to reduce the number of calls.

/cc @rifelpet

[bharath-123]

I was not aware of this validation code. I ll work with this code. I think it makes sense to include during the deep validation phase.

Plus I think the instanceType caching is feature flagged as per https://github.com/kubernetes/kops/blob/master/pkg/featureflag/featureflag.go#L52 ?

[hakman]

Plus I think the instanceType caching is feature flagged as per https://github.com/kubernetes/kops/blob/master/pkg/featureflag/featureflag.go#L52 ?

Yes, this is the PR #9908. I think can be ignored for the initial implementation.

[bharath-123]

@hakman I ll work on moving the validation logic to config validation time. I see the GetMachineTypeInfo

kops/upup/pkg/fi/cloudup/awsup/machine_types.go

Line 69 in d5d08a4

func GetMachineTypeInfo(c AWSCloud, machineType string) (*AWSMachineTypeInfo, error) {

which caches and returns instanceTypeInfo. If I can add a field to cache the architecture field, I think that should be useful over here, don't you think?

[bharath-123]

@olemarkus @rifelpet @hakman Do have a look at this.

[hakman]

Does this make sense anymore? No reason to check the instance type in 2 places.

[bharath-123]

Agreed we can push it to the arch instance type validation function. Will update this

[rifelpet]

We should also check against all instance types in a mixed instance policy in addition to spec.machineType

EDIT: This is already being done

[bharath-123]

I think this should be good for another round of review? @hakman @rifelpet

[rifelpet]

I'm trying to find any validation that ensures InstanceGroup.Spec.Image is not empty but I don't see anything. I'm not sure if that is a valid case, but if the image were to be empty we would no longer be validating the instance type on its own.

[bharath-123]

/retest

[rifelpet]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: bharath-123, rifelpet

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:

• OWNERS [rifelpet]

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