✨ Add the support for capacity blocks #5211

athiruma · 2024-11-08T07:41:22Z

What type of PR is this?
/kind feature

What this PR does / why we need it:
This PR added the support for CapacityBlocks for ML, by adding this we can get the support for capacity reservations of capacity blocks for ML provided by AWS.

Which issue(s) this PR fixes (optional, in fixes #<issue number>(, fixes #<issue_number>, ...) format, will close the issue(s) when PR gets merged):
Fixes # #5045

Special notes for your reviewer:

Checklist:

Release note:

Support running EC2 instances from a reserved capacity block (new API field `marketType`). i.e, marketType: capacity-block

k8s-ci-robot · 2024-11-08T07:41:31Z

Hi @athiruma. Thanks for your PR.

I'm waiting for a kubernetes-sigs 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-sigs/prow repository.

damdo

Thank you for your PR @athiruma 👍
Left some suggestions to improve the overall readability.

pkg/cloud/services/ec2/instances.go

api/v1beta2/awsmachine_types.go

api/v1beta2/types.go

pkg/cloud/services/ec2/instances_test.go

athiruma · 2024-11-12T03:07:42Z

@damdo??

richardcase · 2024-11-12T08:49:52Z

/ok-to-test

athiruma · 2024-11-12T09:21:29Z

/test pull-cluster-api-provider-aws-test

pkg/cloud/services/ec2/launchtemplate.go

damdo · 2024-11-12T10:45:58Z

/test pull-cluster-api-provider-aws-e2e
/test pull-cluster-api-provider-aws-e2e-eks

damdo · 2024-11-12T11:39:08Z

/test pull-cluster-api-provider-aws-e2e

damdo · 2024-11-12T12:19:33Z

@athiruma The pull-cluster-api-provider-aws-e2e is failing on a the CloudFormation timeout.
We've just increased the timeout in #5193 so if you rebase this PR on the latest main you should pick up the bumped timeout and hopefully this will increase the chances of the e2e to pass.

damdo · 2024-11-12T13:05:09Z

/test pull-cluster-api-provider-aws-e2e
/test pull-cluster-api-provider-aws-e2e-eks

athiruma · 2024-11-12T15:55:53Z

/test pull-cluster-api-provider-aws-e2e-eks

athiruma · 2024-11-12T16:19:51Z

/test pull-cluster-api-provider-aws-e2e-eks

damdo · 2024-11-12T19:43:19Z

The eks test is still flakey. I'll take a look at it tomorrow, I think I know where we can improve it

athiruma · 2024-11-14T11:04:38Z

/test pull-cluster-api-provider-aws-e2e-eks

damdo · 2024-11-14T15:29:34Z

@athiruma see #5215 to improve some issues in the EKS tests

athiruma · 2024-12-11T04:39:56Z

@damdo ptal ?

api/v1beta2/awsmachine_types.go

JoelSpeed · 2024-12-18T11:13:21Z

api/v1beta2/awsmachine_types.go

+	//  If this value is selected, CapacityReservationID must be specified to identify the target reservation.
+	// If MarketType is not specified and SpotMarketOptions is provided, the MarketType defaults to "spot".
+	// +optional
+	MarketType *MarketType `json:"marketType,omitempty"`


Does this need to be a pointer?

It should allow empty values so I thought we should use a pointer here.

Do you care about the difference between marketType: "" and the field not being present at all? The controller will treat those two representations in the same way won't it?

Hmm, I see then I'll make a change. Thanks

JoelSpeed · 2024-12-18T11:14:13Z

api/v1beta2/types.go

+const (
+	// MarketTypeOnDemand is a MarketType enum value
+	MarketTypeOnDemand MarketType = "on-demand"
+
+	// MarketTypeSpot is a MarketType enum value
+	MarketTypeSpot MarketType = "spot"
+
+	// MarketTypeCapacityBlock is a MarketType enum value
+	MarketTypeCapacityBlock MarketType = "capacity-block"
+)


Kubernetes API conventions state that enum values should be PascalCase. We prefer this and then to convert to upstream APIs (EC2) in this case so that there's consistency across Kube like APIs. So these should really be OnDemand, Spot and CapacityBlock

api/v1beta2/types.go

enxebre · 2024-12-18T11:55:27Z

api/v1beta2/awsmachine_webhook.go

@@ -361,6 +362,14 @@ func (r *AWSMachine) validateNetworkElasticIPPool() field.ErrorList {
 	return allErrs
 }

+func (r *AWSMachine) validateInstanceMarketType() field.ErrorList {
+	var allErrs field.ErrorList
+	if ptr.Deref(r.Spec.MarketType, "") == MarketTypeCapacityBlock && r.Spec.SpotMarketOptions != nil {


shouldn't this also check MarketType is not ondemand if SpotMarketOptions != nil? i.e. the check should be that the only valid when SpotMarketOptions != nil is MarketType==spot

oh yeah good catch thanks

api/v1beta2/awsmachine_types.go

JoelSpeed

Just some documentation fixes for the new casing and LGTM from the API perspective

JoelSpeed · 2025-01-02T14:24:39Z

api/v1beta2/awsmachine_webhook.go

+func (r *AWSMachine) validateInstanceMarketType() field.ErrorList {
+	var allErrs field.ErrorList
+	if r.Spec.MarketType == MarketTypeCapacityBlock && r.Spec.SpotMarketOptions != nil {
+		allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "MarketType set to CapacityBlock and spotMarketOptions cannot be used together"))


Suggested change

allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "MarketType set to CapacityBlock and spotMarketOptions cannot be used together"))

allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "marketType"), "marketType set to CapacityBlock and spotMarketOptions cannot be used together"))

JoelSpeed · 2025-01-02T14:46:32Z

api/v1beta2/awsmachine_webhook.go

+		allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "MarketType set to CapacityBlock and spotMarketOptions cannot be used together"))
+	}
+	if r.Spec.MarketType == MarketTypeOnDemand && r.Spec.SpotMarketOptions != nil {
+		allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "setting MarketType to OnDemand and spotMarketOptions cannot be used together"))


Suggested change

allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "setting MarketType to OnDemand and spotMarketOptions cannot be used together"))

allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "marketType"), "setting marketType to OnDemand and spotMarketOptions cannot be used together"))

JoelSpeed · 2025-01-02T14:48:33Z

api/v1beta2/types.go

@@ -261,8 +261,32 @@ type Instance struct {
 	// CapacityReservationID specifies the target Capacity Reservation into which the instance should be launched.
 	// +optional
 	CapacityReservationID *string `json:"capacityReservationId,omitempty"`
+
+	// marketType specifies the type of market for the EC2 instance. Valid values include:
+	// "on-demand" (default): The instance runs as a standard On-Demand instance.


Suggested change

// "on-demand" (default): The instance runs as a standard On-Demand instance.

// "OnDemand" (default): The instance runs as a standard On-Demand instance.

JoelSpeed · 2025-01-02T14:48:45Z

api/v1beta2/types.go

+
+	// marketType specifies the type of market for the EC2 instance. Valid values include:
+	// "on-demand" (default): The instance runs as a standard On-Demand instance.
+	// "spot": The instance runs as a Spot instance. When SpotMarketOptions is provided, the MarketType defaults to "spot".


Suggested change

// "spot": The instance runs as a Spot instance. When SpotMarketOptions is provided, the MarketType defaults to "spot".

// "Spot": The instance runs as a Spot instance. When spotMarketOptions is provided, the marketType defaults to "Spot".

JoelSpeed · 2025-01-02T14:49:25Z

api/v1beta2/types.go

+	// marketType specifies the type of market for the EC2 instance. Valid values include:
+	// "on-demand" (default): The instance runs as a standard On-Demand instance.
+	// "spot": The instance runs as a Spot instance. When SpotMarketOptions is provided, the MarketType defaults to "spot".
+	// "capacity-block": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.


Suggested change

// "capacity-block": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.

// "CapacityBlock": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.

JoelSpeed · 2025-01-02T14:49:42Z

api/v1beta2/types.go

+	// "spot": The instance runs as a Spot instance. When SpotMarketOptions is provided, the MarketType defaults to "spot".
+	// "capacity-block": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.
+	//  If this value is selected, CapacityReservationID must be specified to identify the target reservation.
+	// If MarketType is not specified and SpotMarketOptions is provided, the MarketType defaults to "spot".


Suggested change

// If MarketType is not specified and SpotMarketOptions is provided, the MarketType defaults to "spot".

// If marketType is not specified and spotMarketOptions is provided, the marketType defaults to "Spot".

athiruma · 2025-01-02T15:59:15Z

Just some documentation fixes for the new casing and LGTM from the API perspective

fixed

JoelSpeed · 2025-01-02T16:49:57Z

exp/api/v1beta2/types.go

@@ -128,6 +128,19 @@ type AWSLaunchTemplate struct {
 	// PrivateDNSName is the options for the instance hostname.
 	// +optional
 	PrivateDNSName *infrav1.PrivateDNSName `json:"privateDnsName,omitempty"`
+
+	// CapacityReservationID specifies the target Capacity Reservation into which the instance should be launched.


Suggested change

// CapacityReservationID specifies the target Capacity Reservation into which the instance should be launched.

// capacityReservationID specifies the target Capacity Reservation into which the instance should be launched.

Sorry, but this is supposed to be exactly the other way around. Go doc comments are meant to start with the name they describe, as shown for example in the rest of the file. Since the name is capital CapacityReservationID, it's also spelled this way in the comment. Same for all new fields.

That is not true for Kubernetes APIs, it is convention to use the serialized version within the godoc so that commands like kubectl explain use the same language as the end user will be familiar with. Cluster API has also adopted this convention as well, see kubernetes-sigs/cluster-api#11238

I had assumed it was being adopted community wide?

+1 to following the CAPI and Kube conventions. I think Kube's unique among Go projects in how it exposes GoDocs to end users since the type information ends up in the CRD's YAML, but it makes sense to unify the comments so that we're not writing one for developers and one for end users.

I know Joel's written an API linter (https://github.com/JoelSpeed/kal) that he'll be proposing for upstream CAPI. Using this in a GitHub Action would make these conventions easier to follow, I think.

pkg/cloud/services/ec2/instances.go

athiruma · 2025-01-07T07:03:04Z

@JoelSpeed

JoelSpeed

/lgtm

athiruma · 2025-01-09T07:06:20Z

@AndiDog

api/v1beta2/awsmachine_webhook_test.go

exp/api/v1beta2/awsmachinepool_webhook_test.go

AndiDog · 2025-01-09T08:48:43Z

pkg/cloud/services/ec2/instances.go

+		return nil, nil
+	default:
+		// Invalid MarketType provided
+		return nil, errors.Errorf("invalid MarketType %s, must be spot/capacity-block or empty", i.MarketType)


Suggested change

return nil, errors.Errorf("invalid MarketType %s, must be spot/capacity-block or empty", i.MarketType)

return nil, errors.Errorf("invalid MarketType %q", i.MarketType)

AndiDog · 2025-01-09T08:56:01Z

pkg/cloud/services/ec2/launchtemplate.go

+func getLaunchTemplateInstanceMarketOptionsRequest(i *expinfrav1.AWSLaunchTemplate) (*ec2.LaunchTemplateInstanceMarketOptionsRequest, error) {
+	if i.MarketType != "" && i.MarketType != infrav1.MarketTypeCapacityBlock && i.SpotMarketOptions != nil {


This function does the same thing as getInstanceMarketOptionsRequest. Sad that the distinct types don't allow it to easily be one function.

From comparing both functions, I think this is a bug:

Suggested change

func getLaunchTemplateInstanceMarketOptionsRequest(i *expinfrav1.AWSLaunchTemplate) (*ec2.LaunchTemplateInstanceMarketOptionsRequest, error) {

if i.MarketType != "" && i.MarketType != infrav1.MarketTypeCapacityBlock && i.SpotMarketOptions != nil {

func getLaunchTemplateInstanceMarketOptionsRequest(i *expinfrav1.AWSLaunchTemplate) (*ec2.LaunchTemplateInstanceMarketOptionsRequest, error) {

if i.MarketType != "" && i.MarketType == infrav1.MarketTypeCapacityBlock && i.SpotMarketOptions != nil {

ahh, thanks. fixed

AndiDog · 2025-01-09T08:56:27Z

pkg/cloud/services/ec2/launchtemplate_test.go

@@ -1076,6 +1076,7 @@ func TestCreateLaunchTemplateVersion(t *testing.T) {
 		awsResourceReference []infrav1.AWSResourceReference
 		expect               func(m *mocks.MockEC2APIMockRecorder)
 		wantErr              bool
+		MarketType           *string


Suggested change

MarketType *string

marketType *string

k8s-ci-robot · 2025-01-09T09:51:23Z

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please ask for approval from andidog. For more information see the 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:

OWNERS

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

athiruma · 2025-01-10T14:32:23Z

@AndiDog fixed the comments

athiruma · 2025-01-16T11:48:30Z

@AndiDog Can you review the changes?

athiruma · 2025-01-22T15:54:00Z

@JoelSpeed ptal!

JoelSpeed · 2025-01-22T16:13:14Z

The only outstanding thing I had was about the capitalization of the godoc comments, see https://github.com/kubernetes-sigs/cluster-api-provider-aws/pull/5211/files#r1908405924

In core CAPI we agreed to follow the Kube API convention here, I wanted to know if AWS maintainers wanted to follow suit

athiruma · 2025-01-22T16:25:17Z

The only outstanding thing I had was about the capitalization of the godoc comments, see https://github.com/kubernetes-sigs/cluster-api-provider-aws/pull/5211/files#r1908405924

In core CAPI we agreed to follow the Kube API convention here, I wanted to know if AWS maintainers wanted to follow suit

I agree with your point. I think it would be better to address the GoDoc changes in the new PR, as it will make further discussions easier. WDYT?

JoelSpeed · 2025-01-22T16:55:24Z

I'll follow up on the godoc convention separately

/lgtm

nrb · 2025-01-22T16:57:04Z

I think we could do a follow up PR, and just make sure that gets in before we tag a release.

What are your thoughts @AndiDog?

/lgtm

k8s-ci-robot added the size/L Denotes a PR that changes 100-499 lines, ignoring generated files. label Nov 8, 2024

k8s-ci-robot requested review from dlipovetsky and richardcase November 8, 2024 07:41

damdo reviewed Nov 8, 2024

View reviewed changes

athiruma force-pushed the add_instance_market_option branch from 8ae7c7b to d34e59b Compare November 8, 2024 19:16

k8s-ci-robot added size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. and removed size/L Denotes a PR that changes 100-499 lines, ignoring generated files. labels Nov 8, 2024

k8s-ci-robot added ok-to-test Indicates a non-member PR verified by an org member that is safe to test. and removed needs-ok-to-test Indicates a PR that requires an org member to verify it is safe to test. labels Nov 12, 2024

damdo reviewed Nov 12, 2024

View reviewed changes

pkg/cloud/services/ec2/launchtemplate.go Outdated Show resolved Hide resolved

athiruma force-pushed the add_instance_market_option branch from d34e59b to f21fca7 Compare November 12, 2024 09:51

athiruma force-pushed the add_instance_market_option branch from f21fca7 to 05baf2d Compare November 12, 2024 12:52

athiruma force-pushed the add_instance_market_option branch from b5997f8 to 7a2948f Compare December 12, 2024 15:33

JoelSpeed reviewed Dec 18, 2024

View reviewed changes

enxebre reviewed Dec 18, 2024

View reviewed changes

api/v1beta2/awsmachine_types.go Show resolved Hide resolved

athiruma force-pushed the add_instance_market_option branch from 7a2948f to c1d94dc Compare December 20, 2024 15:26

JoelSpeed reviewed Jan 2, 2025

View reviewed changes

athiruma force-pushed the add_instance_market_option branch 2 times, most recently from 8f0b86f to a1fa3fd Compare January 2, 2025 15:58

JoelSpeed reviewed Jan 2, 2025

View reviewed changes

athiruma force-pushed the add_instance_market_option branch from a1fa3fd to af2ddda Compare January 2, 2025 18:06

JoelSpeed reviewed Jan 8, 2025

View reviewed changes

k8s-ci-robot assigned JoelSpeed Jan 8, 2025

k8s-ci-robot added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Jan 8, 2025

AndiDog reviewed Jan 9, 2025

View reviewed changes

Added the support for capacity blocks

66769de

athiruma force-pushed the add_instance_market_option branch from af2ddda to 66769de Compare January 9, 2025 09:51

k8s-ci-robot removed the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Jan 9, 2025

k8s-ci-robot added the lgtm "Looks good to me", indicates that a PR is ready to be merged. label Jan 22, 2025

	allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "MarketType"), "MarketType set to CapacityBlock and spotMarketOptions cannot be used together"))
	allErrs = append(allErrs, field.Forbidden(field.NewPath("spec", "marketType"), "marketType set to CapacityBlock and spotMarketOptions cannot be used together"))

	// "on-demand" (default): The instance runs as a standard On-Demand instance.
	// "OnDemand" (default): The instance runs as a standard On-Demand instance.

	// "spot": The instance runs as a Spot instance. When SpotMarketOptions is provided, the MarketType defaults to "spot".
	// "Spot": The instance runs as a Spot instance. When spotMarketOptions is provided, the marketType defaults to "Spot".

	// "capacity-block": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.
	// "CapacityBlock": The instance utilizes pre-purchased compute capacity (capacity blocks) with AWS Capacity Reservations.

	// If MarketType is not specified and SpotMarketOptions is provided, the MarketType defaults to "spot".
	// If marketType is not specified and spotMarketOptions is provided, the marketType defaults to "Spot".

	// CapacityReservationID specifies the target Capacity Reservation into which the instance should be launched.
	// capacityReservationID specifies the target Capacity Reservation into which the instance should be launched.

	return nil, errors.Errorf("invalid MarketType %s, must be spot/capacity-block or empty", i.MarketType)
	return nil, errors.Errorf("invalid MarketType %q", i.MarketType)

		func getLaunchTemplateInstanceMarketOptionsRequest(i expinfrav1.AWSLaunchTemplate) (ec2.LaunchTemplateInstanceMarketOptionsRequest, error) {
		if i.MarketType != "" && i.MarketType != infrav1.MarketTypeCapacityBlock && i.SpotMarketOptions != nil {

✨ Add the support for capacity blocks #5211

Are you sure you want to change the base?

✨ Add the support for capacity blocks #5211

Conversation

athiruma commented Nov 8, 2024 • edited Loading

k8s-ci-robot commented Nov 8, 2024

damdo left a comment

Choose a reason for hiding this comment

athiruma commented Nov 12, 2024

richardcase commented Nov 12, 2024

athiruma commented Nov 12, 2024

damdo commented Nov 12, 2024

damdo commented Nov 12, 2024

damdo commented Nov 12, 2024

damdo commented Nov 12, 2024

athiruma commented Nov 12, 2024

athiruma commented Nov 12, 2024

damdo commented Nov 12, 2024

athiruma commented Nov 14, 2024

damdo commented Nov 14, 2024

athiruma commented Dec 11, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

enxebre Dec 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

JoelSpeed left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

athiruma commented Jan 2, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

athiruma commented Jan 7, 2025

JoelSpeed left a comment

Choose a reason for hiding this comment

athiruma commented Jan 9, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

k8s-ci-robot commented Jan 9, 2025

athiruma commented Jan 10, 2025 • edited Loading

athiruma commented Jan 16, 2025 • edited Loading

athiruma commented Jan 22, 2025

JoelSpeed commented Jan 22, 2025

athiruma commented Jan 22, 2025

JoelSpeed commented Jan 22, 2025

nrb commented Jan 22, 2025

athiruma commented Nov 8, 2024 •

edited

Loading

athiruma commented Dec 11, 2024 •

edited

Loading

enxebre Dec 18, 2024 •

edited

Loading

athiruma commented Jan 10, 2025 •

edited

Loading

athiruma commented Jan 16, 2025 •

edited

Loading