✨ Remediate unhealthy MachinePool machines

Makefile

@@ -603,6 +603,7 @@ generate-e2e-templates-v1.11: $(KUSTOMIZE)
 generate-e2e-templates-main: $(KUSTOMIZE)
 	$(KUSTOMIZE) build $(DOCKER_TEMPLATES)/main/cluster-template --load-restrictor LoadRestrictionsNone > $(DOCKER_TEMPLATES)/main/cluster-template.yaml
 	$(KUSTOMIZE) build $(DOCKER_TEMPLATES)/main/cluster-template-md-remediation --load-restrictor LoadRestrictionsNone > $(DOCKER_TEMPLATES)/main/cluster-template-md-remediation.yaml
+	$(KUSTOMIZE) build $(DOCKER_TEMPLATES)/main/cluster-template-mp-remediation --load-restrictor LoadRestrictionsNone > $(DOCKER_TEMPLATES)/main/cluster-template-mp-remediation.yaml
 	$(KUSTOMIZE) build $(DOCKER_TEMPLATES)/main/cluster-template-kcp-remediation --load-restrictor LoadRestrictionsNone > $(DOCKER_TEMPLATES)/main/cluster-template-kcp-remediation.yaml
 	$(KUSTOMIZE) build $(DOCKER_TEMPLATES)/main/cluster-template-kcp-adoption/step1 --load-restrictor LoadRestrictionsNone > $(DOCKER_TEMPLATES)/main/cluster-template-kcp-adoption.yaml
 	echo "---" >> $(DOCKER_TEMPLATES)/main/cluster-template-kcp-adoption.yaml

api/core/v1beta1/conversion.go

@@ -1854,6 +1854,10 @@ func Convert_v1beta1_MachinePoolSpec_To_v1beta2_MachinePoolSpec(in *MachinePoolS
 	return autoConvert_v1beta1_MachinePoolSpec_To_v1beta2_MachinePoolSpec(in, out, s)
 }
 
+func Convert_v1beta2_MachinePoolSpec_To_v1beta1_MachinePoolSpec(in *clusterv1.MachinePoolSpec, out *MachinePoolSpec, s apimachineryconversion.Scope) error {
+	return autoConvert_v1beta2_MachinePoolSpec_To_v1beta1_MachinePoolSpec(in, out, s)
+}
+
 func Convert_v1beta1_ClusterClassStatusVariableDefinition_To_v1beta2_ClusterClassStatusVariableDefinition(in *ClusterClassStatusVariableDefinition, out *clusterv1.ClusterClassStatusVariableDefinition, s apimachineryconversion.Scope) error {
 	if err := autoConvert_v1beta1_ClusterClassStatusVariableDefinition_To_v1beta2_ClusterClassStatusVariableDefinition(in, out, s); err != nil {
 		return err

api/core/v1beta1/conversion_test.go

@@ -781,6 +781,7 @@ func spokeObjectReference(in *corev1.ObjectReference, c randfill.Continue) {
 
 func MachinePoolFuzzFuncs(_ runtimeserializer.CodecFactory) []interface{} {
 	return []interface{}{
+		hubMachinePoolSpec,
 		hubMachinePoolStatus,
 		hubMachineSpec,
 		spokeMachinePool,
@@ -789,6 +790,13 @@ func MachinePoolFuzzFuncs(_ runtimeserializer.CodecFactory) []interface{} {
 	}
 }
 
+func hubMachinePoolSpec(in *clusterv1.MachinePoolSpec, c randfill.Continue) {
+	c.FillNoCustom(in)
+
+	// Readded in v1beta2 with different type than in v1alpha's, no conversion possible
+	in.Remediation = clusterv1.MachinePoolRemediationSpec{}
+}
+
 func hubMachinePoolStatus(in *clusterv1.MachinePoolStatus, c randfill.Continue) {
 	c.FillNoCustom(in)
 	// Always create struct with at least one mandatory fields.

api/core/v1beta1/machinepool_types.go

@@ -104,6 +104,17 @@ type MachinePoolSpec struct {
 	FailureDomains []string `json:"failureDomains,omitempty"`
 }
 
+// MachinePoolStrategy describes how to replace existing machines
+// with new ones.
+type MachinePoolStrategy struct {
+	// remediation controls the strategy of remediating unhealthy machines
+	// as marked by a MachineHealthCheck. This only applies to infrastructure
+	// providers supporting "MachinePool Machines". For other providers,
+	// no remediation is done.
+	// +optional
+	Remediation *RemediationStrategy `json:"remediation,omitempty"`
+}
+
 // MachinePoolStatus defines the observed state of MachinePool.
 type MachinePoolStatus struct {
 	// nodeRefs will point to the corresponding Nodes if it they exist.

api/core/v1beta2/machinepool_types.go

@@ -19,6 +19,7 @@ package v1beta2
 import (
 	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/intstr"
 
 	capierrors "sigs.k8s.io/cluster-api/errors"
 )
@@ -97,6 +98,29 @@ type MachinePoolSpec struct {
 	// +kubebuilder:validation:items:MinLength=1
 	// +kubebuilder:validation:items:MaxLength=256
 	FailureDomains []string `json:"failureDomains,omitempty"`
+
+	// remediation controls how unhealthy Machines are remediated (through a MachineHealthCheck).
+	// This only applies to infrastructure providers supporting and enabling the
+	// "MachinePool Machines" feature. For other setups, no remediation is done.
+	// +optional
+	Remediation MachinePoolRemediationSpec `json:"remediation,omitempty,omitzero"`
+}
+
+// MachinePoolRemediationSpec controls how unhealthy Machines are remediated (through a MachineHealthCheck).
+// This only applies to infrastructure providers supporting and enabling the
+// "MachinePool Machines" feature. For other setups, no remediation is done.
+// +kubebuilder:validation:MinProperties=1
+type MachinePoolRemediationSpec struct {
+	// maxInFlight determines how many in flight remediations should happen at the same time.
+	//
+	// MaxInFlight can be set to a fixed number or a percentage.
+	// Example: when this is set to 20%, the MachinePool controller deletes at most 20% of
+	// the desired replicas.
+	//
+	// If not set, remediation is limited to all machines under the active MachinePool's management.
+	//
+	// +optional
+	MaxInFlight *intstr.IntOrString `json:"maxInFlight,omitempty"`
 }
 
 // MachinePoolStatus defines the observed state of MachinePool.

go.mod

@@ -34,7 +34,7 @@ require (
 	go.etcd.io/etcd/client/pkg/v3 v3.6.5
 	go.etcd.io/etcd/client/v3 v3.6.5
 	go.uber.org/zap v1.27.0
-	golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56 // indirect
+	golang.org/x/exp v0.0.0-20240719175910-8a7402abbf56
 	golang.org/x/oauth2 v0.32.0
 	golang.org/x/text v0.30.0
 	gomodules.xyz/jsonpatch/v2 v2.5.0

internal/api/core/v1alpha3/conversion.go

@@ -655,6 +655,16 @@ func (dst *MachinePool) ConvertFrom(srcRaw conversion.Hub) error {
 	return utilconversion.MarshalData(src, dst)
 }
 
+func Convert_v1alpha3_MachineDeploymentStrategy_To_v1beta2_MachinePoolStrategy(in *MachineDeploymentStrategy, out *clusterv1.MachinePoolRemediationSpec, _ apimachineryconversion.Scope) error {
+	// Subfields differ in v1beta2, no conversion possible
+	out.MaxInFlight = nil
+	return nil
+}
+
+func Convert_v1beta2_MachinePoolStrategy_To_v1alpha3_MachineDeploymentStrategy(in *clusterv1.MachinePoolRemediationSpec, out *MachineDeploymentStrategy, _ apimachineryconversion.Scope) error {
+	return nil
+}
+
 func Convert_v1beta2_MachineSetStatus_To_v1alpha3_MachineSetStatus(in *clusterv1.MachineSetStatus, out *MachineSetStatus, _ apimachineryconversion.Scope) error {
 	// Status.Conditions was introduced in v1alpha4, thus requiring a custom conversion function; the values is going to be preserved in an annotation thus allowing roundtrip without loosing informations
 	// V1Beta2 was added in v1beta1.
@@ -973,7 +983,10 @@ func Convert_v1alpha3_MachineSetSpec_To_v1beta2_MachineSetSpec(in *MachineSetSpe
 	return nil
 }
 
-// Convert_v1alpha3_MachinePoolSpec_To_v1beta2_MachinePoolSpec is an autogenerated conversion function.
+func Convert_v1beta2_MachinePoolSpec_To_v1alpha3_MachinePoolSpec(in *clusterv1.MachinePoolSpec, out *MachinePoolSpec, s apimachineryconversion.Scope) error {
+	return autoConvert_v1beta2_MachinePoolSpec_To_v1alpha3_MachinePoolSpec(in, out, s)
+}
+
 func Convert_v1alpha3_MachinePoolSpec_To_v1beta2_MachinePoolSpec(in *MachinePoolSpec, out *clusterv1.MachinePoolSpec, s apimachineryconversion.Scope) error {
 	return autoConvert_v1alpha3_MachinePoolSpec_To_v1beta2_MachinePoolSpec(in, out, s)
 }