✨ MachineHealthCheck supports checking Machine conditions

api/core/v1beta1/conversion_test.go

@@ -75,6 +75,7 @@ func TestFuzzyConversion(t *testing.T) {
 
 func ClusterFuzzFuncs(_ runtimeserializer.CodecFactory) []interface{} {
 	return []interface{}{
+		hubClusterSpec,
 		hubClusterStatus,
 		spokeClusterTopology,
 		spokeClusterStatus,
@@ -99,6 +100,21 @@ func hubClusterStatus(in *clusterv1.ClusterStatus, c randfill.Continue) {
 	}
 }
 
+func hubClusterSpec(in *clusterv1.ClusterSpec, c randfill.Continue) {
+	c.FillNoCustom(in)
+
+	// remove MachineHealthCheck.UnhealthyMachineConditions as it does not exist in v1beta1.
+	if in.Topology != nil && in.Topology.ControlPlane.MachineHealthCheck != nil {
+		in.Topology.ControlPlane.MachineHealthCheck.UnhealthyMachineConditions = nil
+	}
+
+	if in.Topology != nil && in.Topology.Workers != nil && len(in.Topology.Workers.MachineDeployments) > 0 {
+		for i := range in.Topology.Workers.MachineDeployments {
+			in.Topology.Workers.MachineDeployments[i].MachineHealthCheck = nil
+		}
+	}
+}
+
 func spokeClusterTopology(in *Topology, c randfill.Continue) {
 	c.FillNoCustom(in)
 
@@ -129,6 +145,22 @@ func ClusterClassFuncs(_ runtimeserializer.CodecFactory) []interface{} {
 		hubJSONSchemaProps,
 		spokeClusterClassStatus,
 		spokeJSONSchemaProps,
+		hubClusterClassSpec,
+	}
+}
+
+func hubClusterClassSpec(in *clusterv1.ClusterClassSpec, c randfill.Continue) {
+	c.FillNoCustom(in)
+
+	// remove MachineHealthCheck.UnhealthyMachineConditions as it does not exist in v1beta1.
+	if in.ControlPlane.MachineHealthCheck != nil && in.ControlPlane.MachineHealthCheck.UnhealthyMachineConditions != nil {
+		in.ControlPlane.MachineHealthCheck.UnhealthyMachineConditions = nil
+	}
+
+	if len(in.Workers.MachineDeployments) > 0 {
+		for i := range in.Workers.MachineDeployments {
+			in.Workers.MachineDeployments[i].MachineHealthCheck = nil
+		}
 	}
 }
 
@@ -348,13 +380,24 @@ func spokeMachineDeploymentStatus(in *MachineDeploymentStatus, c randfill.Contin
 
 func MachineHealthCheckFuzzFuncs(_ runtimeserializer.CodecFactory) []interface{} {
 	return []interface{}{
+		hubMachineHealthCheckSpec,
 		hubMachineHealthCheckStatus,
 		spokeMachineHealthCheckStatus,
 	}
 }
 
+func hubMachineHealthCheckSpec(in *clusterv1.MachineHealthCheckSpec, c randfill.Continue) {
+	c.FillNoCustom(in)
+
+	// Drop UnhealthyMachineConditions as it does not exist in v1beta1.
+	if in.UnhealthyMachineConditions != nil {
+		in.UnhealthyMachineConditions = nil
+	}
+}
+
 func hubMachineHealthCheckStatus(in *clusterv1.MachineHealthCheckStatus, c randfill.Continue) {
 	c.FillNoCustom(in)
+
 	// Drop empty structs with only omit empty fields.
 	if in.Deprecated != nil {
 		if in.Deprecated.V1Beta1 == nil || reflect.DeepEqual(in.Deprecated.V1Beta1, &clusterv1.MachineHealthCheckV1Beta1DeprecatedStatus{}) {

api/core/v1beta2/clusterclass_types.go

@@ -385,6 +385,14 @@ type MachineHealthCheckClass struct {
 	// +kubebuilder:validation:MaxItems=100
 	UnhealthyNodeConditions []UnhealthyNodeCondition `json:"unhealthyNodeConditions,omitempty"`
 
+	// unhealthyMachineConditions contains a list of the machine conditions that determine
+	// whether a node is considered unhealthy.  The conditions are combined in a
+	// logical OR, i.e. if any of the conditions is met, the node is unhealthy.
+	//
+	// +optional
+	// +kubebuilder:validation:MaxItems=100
+	UnhealthyMachineConditions []UnhealthyMachineCondition `json:"unhealthyMachineConditions,omitempty"`
+
 	// maxUnhealthy specifies the maximum number of unhealthy machines allowed.
 	// Any further remediation is only allowed if at most "maxUnhealthy" machines selected by
 	// "selector" are not healthy.

api/core/v1beta2/machine_types.go

@@ -276,6 +276,10 @@ const (
 	// defined by a MachineHealthCheck object.
 	MachineHealthCheckUnhealthyNodeReason = "UnhealthyNode"
 
+	// MachineHealthCheckUnhealthyMachineReason surfaces when the machine does not pass the health checks
+	// defined by a MachineHealthCheck object.
+	MachineHealthCheckUnhealthyMachineReason = "UnhealthyMachine"
+
 	// MachineHealthCheckNodeStartupTimeoutReason surfaces when the node hosted on the machine does not appear within
 	// the timeout defined by a MachineHealthCheck object.
 	MachineHealthCheckNodeStartupTimeoutReason = "NodeStartupTimeout"

api/core/v1beta2/machinehealthcheck_types.go

@@ -69,6 +69,14 @@ type MachineHealthCheckSpec struct {
 	// +kubebuilder:validation:MaxItems=100
 	UnhealthyNodeConditions []UnhealthyNodeCondition `json:"unhealthyNodeConditions,omitempty"`
 
+	// unhealthyMachineConditions contains a list of the machine conditions that determine
+	// whether a node is considered unhealthy.  The conditions are combined in a
+	// logical OR, i.e. if any of the conditions is met, the node is unhealthy.
+	//
+	// +optional
+	// +kubebuilder:validation:MaxItems=100
+	UnhealthyMachineConditions []UnhealthyMachineCondition `json:"unhealthyMachineConditions,omitempty"`
+
 	// maxUnhealthy specifies the maximum number of unhealthy machines allowed.
 	// Any further remediation is only allowed if at most "maxUnhealthy" machines selected by
 	// "selector" are not healthy.
@@ -148,6 +156,34 @@ type UnhealthyNodeCondition struct {
 
 // ANCHOR_END: UnhealthyNodeCondition
 
+// ANCHOR: UnhealthyMachineCondition
+
+// UnhealthyMachineCondition represents a Node condition type and value with a timeout
+// specified as a duration.  When the named condition has been in the given
+// status for at least the timeout value, a node is considered unhealthy.
+type UnhealthyMachineCondition struct {
+	// type of Node condition
+	// +kubebuilder:validation:Type=string
+	// +kubebuilder:validation:MinLength=1
+	// +kubebuilder:validation:MaxLength=316
+	// +required
+	Type string `json:"type"`
+
+	// status of the condition, one of True, False, Unknown.
+	// +required
+	// +kubebuilder:validation:Enum=True;False;Unknown
+	Status metav1.ConditionStatus `json:"status"`
+
+	// timeout is the duration that a node must be in a given status for,
+	// after which the node is considered unhealthy.
+	// For example, with a value of "1h", the node must match the status
+	// for at least 1 hour before being considered unhealthy.
+	// +required
+	Timeout metav1.Duration `json:"timeout"`
+}
+
+// ANCHOR_END: UnhealthyMachineCondition
+
 // ANCHOR: MachineHealthCheckStatus
 
 // MachineHealthCheckStatus defines the observed state of MachineHealthCheck.

api/core/v1beta2/v1beta1_condition_consts.go

@@ -161,6 +161,9 @@ const (
 
 	// UnhealthyNodeConditionV1Beta1Reason is the reason used when a machine's node has one of the MachineHealthCheck's unhealthy conditions.
 	UnhealthyNodeConditionV1Beta1Reason = "UnhealthyNode"
+
+	// UnhealthyMachineConditionV1Beta1Reason is the reason used when a machine has one of the MachineHealthCheck's unhealthy conditions.
+	UnhealthyMachineConditionV1Beta1Reason = "UnhealthyMachine"
 )
 
 const (