scale.go

/*
Copyright 2020 The Kubernetes Authors.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
*/

package controllers

import (
	"context"
	"sigs.k8s.io/cluster-api/util/collections"
	"strings"

	"github.com/pkg/errors"
	corev1 "k8s.io/api/core/v1"
	apierrors "k8s.io/apimachinery/pkg/api/errors"
	kerrors "k8s.io/apimachinery/pkg/util/errors"
	clusterv1 "sigs.k8s.io/cluster-api/api/v1alpha4"
	controlplanev1 "sigs.k8s.io/cluster-api/controlplane/kubeadm/api/v1alpha4"
	"sigs.k8s.io/cluster-api/controlplane/kubeadm/internal"
	"sigs.k8s.io/cluster-api/util"
	"sigs.k8s.io/cluster-api/util/conditions"
	ctrl "sigs.k8s.io/controller-runtime"
)

func (r *KubeadmControlPlaneReconciler) initializeControlPlane(ctx context.Context, cluster *clusterv1.Cluster, kcp *controlplanev1.KubeadmControlPlane, controlPlane *internal.ControlPlane) (ctrl.Result, error) {
	logger := controlPlane.Logger()

	// Perform an uncached read of all the owned machines. This check is in place to make sure
	// that the controller cache is not misbehaving and we end up initializing the cluster more than once.
	ownedMachines, err := r.managementClusterUncached.GetMachinesForCluster(ctx, cluster, collections.OwnedMachines(kcp))
	if err != nil {
		logger.Error(err, "failed to perform an uncached read of control plane machines for cluster")
		return ctrl.Result{}, err
	}
	if len(ownedMachines) > 0 {
		return ctrl.Result{}, errors.Errorf(
			"control plane has already been initialized, found %d owned machine for cluster %s/%s: controller cache or management cluster is misbehaving",
			len(ownedMachines), cluster.Namespace, cluster.Name,
		)
	}

	bootstrapSpec := controlPlane.InitialControlPlaneConfig()
	fd := controlPlane.NextFailureDomainForScaleUp()
	if err := r.cloneConfigsAndGenerateMachine(ctx, cluster, kcp, bootstrapSpec, fd); err != nil {
		logger.Error(err, "Failed to create initial control plane Machine")
		r.recorder.Eventf(kcp, corev1.EventTypeWarning, "FailedInitialization", "Failed to create initial control plane Machine for cluster %s/%s control plane: %v", cluster.Namespace, cluster.Name, err)
		return ctrl.Result{}, err
	}

	// Requeue the control plane, in case there are additional operations to perform
	return ctrl.Result{Requeue: true}, nil
}

func (r *KubeadmControlPlaneReconciler) scaleUpControlPlane(ctx context.Context, cluster *clusterv1.Cluster, kcp *controlplanev1.KubeadmControlPlane, controlPlane *internal.ControlPlane) (ctrl.Result, error) {
	logger := controlPlane.Logger()

	// Run preflight checks to ensure that the control plane is stable before proceeding with a scale up/scale down operation; if not, wait.
	if result, err := r.preflightChecks(ctx, controlPlane); err != nil || !result.IsZero() {
		return result, err
	}

	// Create the bootstrap configuration
	bootstrapSpec := controlPlane.JoinControlPlaneConfig()
	fd := controlPlane.NextFailureDomainForScaleUp()
	if err := r.cloneConfigsAndGenerateMachine(ctx, cluster, kcp, bootstrapSpec, fd); err != nil {
		logger.Error(err, "Failed to create additional control plane Machine")
		r.recorder.Eventf(kcp, corev1.EventTypeWarning, "FailedScaleUp", "Failed to create additional control plane Machine for cluster %s/%s control plane: %v", cluster.Namespace, cluster.Name, err)
		return ctrl.Result{}, err
	}

	// Requeue the control plane, in case there are other operations to perform
	return ctrl.Result{Requeue: true}, nil
}

func (r *KubeadmControlPlaneReconciler) scaleDownControlPlane(
	ctx context.Context,
	cluster *clusterv1.Cluster,
	kcp *controlplanev1.KubeadmControlPlane,
	controlPlane *internal.ControlPlane,
	outdatedMachines collections.Machines,
) (ctrl.Result, error) {
	logger := controlPlane.Logger()

	// Pick the Machine that we should scale down.
	machineToDelete, err := selectMachineForScaleDown(controlPlane, outdatedMachines)
	if err != nil {
		return ctrl.Result{}, errors.Wrap(err, "failed to select machine for scale down")
	}

	// Run preflight checks ensuring the control plane is stable before proceeding with a scale up/scale down operation; if not, wait.
	// Given that we're scaling down, we can exclude the machineToDelete from the preflight checks.
	if result, err := r.preflightChecks(ctx, controlPlane, machineToDelete); err != nil || !result.IsZero() {
		return result, err
	}

	workloadCluster, err := r.managementCluster.GetWorkloadCluster(ctx, util.ObjectKey(cluster))
	if err != nil {
		logger.Error(err, "Failed to create client to workload cluster")
		return ctrl.Result{}, errors.Wrapf(err, "failed to create client to workload cluster")
	}

	if machineToDelete == nil {
		logger.Info("Failed to pick control plane Machine to delete")
		return ctrl.Result{}, errors.New("failed to pick control plane Machine to delete")
	}

	// If KCP should manage etcd, If etcd leadership is on machine that is about to be deleted, move it to the newest member available.
	if controlPlane.IsEtcdManaged() {
		etcdLeaderCandidate := controlPlane.Machines.Newest()
		if err := workloadCluster.ForwardEtcdLeadership(ctx, machineToDelete, etcdLeaderCandidate); err != nil {
			logger.Error(err, "Failed to move leadership to candidate machine", "candidate", etcdLeaderCandidate.Name)
			return ctrl.Result{}, err
		}
		if err := workloadCluster.RemoveEtcdMemberForMachine(ctx, machineToDelete); err != nil {
			logger.Error(err, "Failed to remove etcd member for machine")
			return ctrl.Result{}, err
		}
	}

	if err := workloadCluster.RemoveMachineFromKubeadmConfigMap(ctx, machineToDelete); err != nil {
		logger.Error(err, "Failed to remove machine from kubeadm ConfigMap")
		return ctrl.Result{}, err
	}

	logger = logger.WithValues("machine", machineToDelete)
	if err := r.Client.Delete(ctx, machineToDelete); err != nil && !apierrors.IsNotFound(err) {
		logger.Error(err, "Failed to delete control plane machine")
		r.recorder.Eventf(kcp, corev1.EventTypeWarning, "FailedScaleDown",
			"Failed to delete control plane Machine %s for cluster %s/%s control plane: %v", machineToDelete.Name, cluster.Namespace, cluster.Name, err)
		return ctrl.Result{}, err
	}

	// Requeue the control plane, in case there are additional operations to perform
	return ctrl.Result{Requeue: true}, nil
}

// preflightChecks checks if the control plane is stable before proceeding with a scale up/scale down operation,
// where stable means that:
// - There are no machine deletion in progress
// - All the health conditions on KCP are true.
// - All the health conditions on the control plane machines are true.
// If the control plane is not passing preflight checks, it requeue.
//
// NOTE: this func uses KCP conditions, it is required to call reconcileControlPlaneConditions before this.
func (r *KubeadmControlPlaneReconciler) preflightChecks(_ context.Context, controlPlane *internal.ControlPlane, excludeFor ...*clusterv1.Machine) (ctrl.Result, error) { //nolint:unparam
	logger := controlPlane.Logger()

	// If there is no KCP-owned control-plane machines, then control-plane has not been initialized yet,
	// so it is considered ok to proceed.
	if controlPlane.Machines.Len() == 0 {
		return ctrl.Result{}, nil
	}

	// If there are deleting machines, wait for the operation to complete.
	if controlPlane.HasDeletingMachine() {
		logger.Info("Waiting for machines to be deleted", "Machines", strings.Join(controlPlane.Machines.Filter(collections.HasDeletionTimestamp).Names(), ", "))
		return ctrl.Result{RequeueAfter: deleteRequeueAfter}, nil
	}

	// Check machine health conditions; if there are conditions with False or Unknown, then wait.
	allMachineHealthConditions := []clusterv1.ConditionType{
		controlplanev1.MachineAPIServerPodHealthyCondition,
		controlplanev1.MachineControllerManagerPodHealthyCondition,
		controlplanev1.MachineSchedulerPodHealthyCondition,
	}
	if controlPlane.IsEtcdManaged() {
		allMachineHealthConditions = append(allMachineHealthConditions,
			controlplanev1.MachineEtcdPodHealthyCondition,
			controlplanev1.MachineEtcdMemberHealthyCondition,
		)
	}
	machineErrors := []error{}

loopmachines:
	for _, machine := range controlPlane.Machines {
		for _, excluded := range excludeFor {
			// If this machine should be excluded from the individual
			// health check, continue the out loop.
			if machine.Name == excluded.Name {
				continue loopmachines
			}
		}

		for _, condition := range allMachineHealthConditions {
			if err := preflightCheckCondition("machine", machine, condition); err != nil {
				machineErrors = append(machineErrors, err)
			}
		}
	}
	if len(machineErrors) > 0 {
		aggregatedError := kerrors.NewAggregate(machineErrors)
		r.recorder.Eventf(controlPlane.KCP, corev1.EventTypeWarning, "ControlPlaneUnhealthy",
			"Waiting for control plane to pass preflight checks to continue reconciliation: %v", aggregatedError)
		logger.Info("Waiting for control plane to pass preflight checks", "failures", aggregatedError.Error())

		return ctrl.Result{RequeueAfter: preflightFailedRequeueAfter}, nil
	}

	return ctrl.Result{}, nil
}

func preflightCheckCondition(kind string, obj conditions.Getter, condition clusterv1.ConditionType) error {
	c := conditions.Get(obj, condition)
	if c == nil {
		return errors.Errorf("%s %s does not have %s condition", kind, obj.GetName(), condition)
	}
	if c.Status == corev1.ConditionFalse {
		return errors.Errorf("%s %s reports %s condition is false (%s, %s)", kind, obj.GetName(), condition, c.Severity, c.Message)
	}
	if c.Status == corev1.ConditionUnknown {
		return errors.Errorf("%s %s reports %s condition is unknown (%s)", kind, obj.GetName(), condition, c.Message)
	}
	return nil
}

func selectMachineForScaleDown(controlPlane *internal.ControlPlane, outdatedMachines collections.Machines) (*clusterv1.Machine, error) {
	machines := controlPlane.Machines
	switch {
	case controlPlane.MachineWithDeleteAnnotation(outdatedMachines).Len() > 0:
		machines = controlPlane.MachineWithDeleteAnnotation(outdatedMachines)
	case controlPlane.MachineWithDeleteAnnotation(machines).Len() > 0:
		machines = controlPlane.MachineWithDeleteAnnotation(machines)
	case outdatedMachines.Len() > 0:
		machines = outdatedMachines
	}
	return controlPlane.MachineInFailureDomainWithMostMachines(machines)
}