🌱 Implement grace period for KCP remote conditions (#11339)

* Implement grace period for KCP remote conditions

Signed-off-by: Stefan Büringer [email protected]

* Fix review findings

* Fix review findings

* Fix review findings

---------

Signed-off-by: Stefan Büringer [email protected]

Author: sbueringer

api/v1beta1/machine_types.go

@@ -206,15 +206,11 @@ const (
 	// during the deletion workflow, or by a users.
 	MachineNodeDeletedV1Beta2Reason = ObjectDeletedV1Beta2Reason
 
-	// MachineNodeRemoteConnectionFailedV1Beta2Reason surfaces that the remote connection failed.
-	// If the remote connection probe failed for longer than remote conditions grace period,
-	// this reason is used when setting NodeHealthy and NodeReady conditions to `Unknown`.
-	MachineNodeRemoteConnectionFailedV1Beta2Reason = RemoteConnectionFailedV1Beta2Reason
-
-	// MachineNodeRemoteConnectionDownV1Beta2Reason surfaces that the remote connection is down.
-	// This is used when setting NodeHealthy and NodeReady conditions to `Unknown`
-	// when the connection is down and they haven't been set yet.
-	MachineNodeRemoteConnectionDownV1Beta2Reason = RemoteConnectionDownV1Beta2Reason
+	// MachineNodeInspectionFailedV1Beta2Reason documents a failure when inspecting the status of a Node.
+	MachineNodeInspectionFailedV1Beta2Reason = InspectionFailedV1Beta2Reason
+
+	// MachineNodeConnectionDownV1Beta2Reason surfaces that the connection to the workload cluster is down.
+	MachineNodeConnectionDownV1Beta2Reason = ConnectionDownV1Beta2Reason
 )
 
 // Machine's HealthCheckSucceeded condition and corresponding reasons that will be used in v1Beta2 API version.

api/v1beta1/v1beta2_condition_consts.go

@@ -106,8 +106,8 @@ const (
 	// set to false and with the OwnerRemediated condition set to false by the MachineHealthCheck controller.
 	RemediatingV1Beta2Reason = "Remediating"
 
-	// NotRemediatingV1Beta2Reason surfaces when an object does not own any machines marked as not healthy
-	// by the MachineHealthCheck controller.
+	// NotRemediatingV1Beta2Reason surfaces when an object does not own any machines with HealthCheckSucceeded
+	// set to false and with the OwnerRemediated condition set to false by the MachineHealthCheck controller.
 	NotRemediatingV1Beta2Reason = "NotRemediating"
 
 	// NoReplicasV1Beta2Reason surfaces when an object that manage replicas does not have any.
@@ -142,15 +142,8 @@ const (
 	// PausedV1Beta2Reason surfaces when an object is paused.
 	PausedV1Beta2Reason = "Paused"
 
-	// RemoteConnectionFailedV1Beta2Reason surfaces that the remote connection failed.
-	// This is typically used when setting remote conditions (e.g. `NodeHealthy`) to `Unknown`
-	// after the remote connection probe didn't succeed for remote conditions grace period.
-	RemoteConnectionFailedV1Beta2Reason = "RemoteConnectionFailed"
-
-	// RemoteConnectionDownV1Beta2Reason surfaces that the remote connection is down.
-	// This is typically used when setting remote conditions (e.g. `NodeHealthy`) to `Unknown`
-	// when the connection is down and they haven't been set yet.
-	RemoteConnectionDownV1Beta2Reason = "RemoteConnectionDown"
+	// ConnectionDownV1Beta2Reason surfaces that the connection to the workload cluster is down.
+	ConnectionDownV1Beta2Reason = "ConnectionDown"
 
 	// DeletionTimestampNotSetV1Beta2Reason surfaces when an object is not deleting because the
 	// DeletionTimestamp is not set.

controlplane/kubeadm/api/v1beta1/v1beta2_condition_consts.go

@@ -20,11 +20,25 @@ import clusterv1 "sigs.k8s.io/cluster-api/api/v1beta1"
 
 // KubeadmControlPlane's Available condition and corresponding reasons that will be used in v1Beta2 API version.
 const (
-	// KubeadmControlPlaneAvailableV1Beta2Condition True if the control plane can be reached, EtcdClusterHealthy is true,
+	// KubeadmControlPlaneAvailableV1Beta2Condition is True if the control plane can be reached, EtcdClusterHealthy is true,
 	// and CertificatesAvailable is true.
 	KubeadmControlPlaneAvailableV1Beta2Condition = clusterv1.AvailableV1Beta2Condition
 )
 
+// KubeadmControlPlane's Initialized condition and corresponding reasons that will be used in v1Beta2 API version.
+const (
+	// KubeadmControlPlaneInitializedV1Beta2Condition is True when the control plane is functional enough to accept
+	// requests. This information is usually used as a signal for starting all the provisioning operations that
+	// depend on a functional API server, but do not require a full HA control plane to exist.
+	KubeadmControlPlaneInitializedV1Beta2Condition = "Initialized"
+
+	// KubeadmControlPlaneInitializedV1Beta2Reason surfaces when the control plane is initialized.
+	KubeadmControlPlaneInitializedV1Beta2Reason = "Initialized"
+
+	// KubeadmControlPlaneNotInitializedV1Beta2Reason surfaces when the control plane is not initialized.
+	KubeadmControlPlaneNotInitializedV1Beta2Reason = "NotInitialized"
+)
+
 // KubeadmControlPlane's CertificatesAvailable condition and corresponding reasons that will be used in v1Beta2 API version.
 const (
 	// KubeadmControlPlaneCertificatesAvailableV1Beta2Condition True if all the cluster certificates exist.
@@ -52,6 +66,10 @@ const (
 	// etcd cluster hosted on KubeadmControlPlane controlled machines.
 	KubeadmControlPlaneEtcdClusterInspectionFailedV1Beta2Reason = clusterv1.InspectionFailedV1Beta2Reason
 
+	// KubeadmControlPlaneEtcdClusterConnectionDownV1Beta2Reason surfaces that the connection to the workload
+	// cluster is down.
+	KubeadmControlPlaneEtcdClusterConnectionDownV1Beta2Reason = clusterv1.ConnectionDownV1Beta2Reason
+
 	// KubeadmControlPlaneEtcdClusterHealthyV1Beta2Reason surfaces when the etcd cluster hosted on KubeadmControlPlane
 	// machines is healthy.
 	KubeadmControlPlaneEtcdClusterHealthyV1Beta2Reason = "Healthy"
@@ -77,6 +95,10 @@ const (
 	// control plane components hosted on KubeadmControlPlane controlled machines.
 	KubeadmControlPlaneControlPlaneComponentsInspectionFailedV1Beta2Reason = clusterv1.InspectionFailedV1Beta2Reason
 
+	// KubeadmControlPlaneControlPlaneComponentsConnectionDownV1Beta2Reason surfaces that the connection to the workload
+	// cluster is down.
+	KubeadmControlPlaneControlPlaneComponentsConnectionDownV1Beta2Reason = clusterv1.ConnectionDownV1Beta2Reason
+
 	// KubeadmControlPlaneControlPlaneComponentsHealthyV1Beta2Reason surfaces when the Kubernetes control plane components
 	// hosted on KubeadmControlPlane machines are healthy.
 	KubeadmControlPlaneControlPlaneComponentsHealthyV1Beta2Reason = "Healthy"
@@ -233,13 +255,13 @@ const (
 	// pod hosted on a KubeadmControlPlane controlled machine.
 	KubeadmControlPlaneMachinePodInspectionFailedV1Beta2Reason = clusterv1.InspectionFailedV1Beta2Reason
 
+	// KubeadmControlPlaneMachinePodConnectionDownV1Beta2Reason surfaces that the connection to the workload
+	// cluster is down.
+	KubeadmControlPlaneMachinePodConnectionDownV1Beta2Reason = clusterv1.ConnectionDownV1Beta2Reason
+
 	// KubeadmControlPlaneMachinePodDeletingV1Beta2Reason surfaces when the machine hosting control plane components
 	// is being deleted.
 	KubeadmControlPlaneMachinePodDeletingV1Beta2Reason = "Deleting"
-
-	// KubeadmControlPlaneMachinePodInternalErrorV1Beta2Reason surfaces unexpected failures when reading pod hosted
-	// on a KubeadmControlPlane controlled machine.
-	KubeadmControlPlaneMachinePodInternalErrorV1Beta2Reason = clusterv1.InternalErrorV1Beta2Reason
 )
 
 // EtcdMemberHealthy condition and corresponding reasons that will be used for KubeadmControlPlane controlled machines in v1Beta2 API version.
@@ -257,6 +279,10 @@ const (
 	// etcd member hosted on a KubeadmControlPlane controlled machine.
 	KubeadmControlPlaneMachineEtcdMemberInspectionFailedV1Beta2Reason = clusterv1.InspectionFailedV1Beta2Reason
 
+	// KubeadmControlPlaneMachineEtcdMemberConnectionDownV1Beta2Reason surfaces that the connection to the workload
+	// cluster is down.
+	KubeadmControlPlaneMachineEtcdMemberConnectionDownV1Beta2Reason = clusterv1.ConnectionDownV1Beta2Reason
+
 	// KubeadmControlPlaneMachineEtcdMemberDeletingV1Beta2Reason surfaces when the machine hosting an etcd member
 	// is being deleted.
 	KubeadmControlPlaneMachineEtcdMemberDeletingV1Beta2Reason = "Deleting"

controlplane/kubeadm/controllers/alias.go

@@ -40,6 +40,8 @@ type KubeadmControlPlaneReconciler struct {
 	// WatchFilterValue is the label value used to filter events prior to reconciliation.
 	WatchFilterValue string
 
+	RemoteConditionsGracePeriod time.Duration
+
 	// Deprecated: DeprecatedInfraMachineNaming. Name the InfraStructureMachines after the InfraMachineTemplate.
 	DeprecatedInfraMachineNaming bool
 }
@@ -53,6 +55,7 @@ func (r *KubeadmControlPlaneReconciler) SetupWithManager(ctx context.Context, mg
 		EtcdDialTimeout:              r.EtcdDialTimeout,
 		EtcdCallTimeout:              r.EtcdCallTimeout,
 		WatchFilterValue:             r.WatchFilterValue,
+		RemoteConditionsGracePeriod:  r.RemoteConditionsGracePeriod,
 		DeprecatedInfraMachineNaming: r.DeprecatedInfraMachineNaming,
 	}).SetupWithManager(ctx, mgr, options)
 }

controlplane/kubeadm/internal/control_plane.go

@@ -309,6 +309,12 @@ func (c *ControlPlane) PatchMachines(ctx context.Context) error {
 				controlplanev1.MachineSchedulerPodHealthyCondition,
 				controlplanev1.MachineEtcdPodHealthyCondition,
 				controlplanev1.MachineEtcdMemberHealthyCondition,
+			}}, patch.WithOwnedV1Beta2Conditions{Conditions: []string{
+				controlplanev1.KubeadmControlPlaneMachineAPIServerPodHealthyV1Beta2Condition,
+				controlplanev1.KubeadmControlPlaneMachineControllerManagerPodHealthyV1Beta2Condition,
+				controlplanev1.KubeadmControlPlaneMachineSchedulerPodHealthyV1Beta2Condition,
+				controlplanev1.KubeadmControlPlaneMachineEtcdPodHealthyV1Beta2Condition,
+				controlplanev1.KubeadmControlPlaneMachineEtcdMemberHealthyV1Beta2Condition,
 			}}); err != nil {
 				errList = append(errList, err)
 			}