cluster.go

/*
Copyright 2021 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 webhooks

import (
	"context"
	"fmt"
	"strings"

	"github.com/blang/semver"
	"github.com/pkg/errors"
	apierrors "k8s.io/apimachinery/pkg/api/errors"
	"k8s.io/apimachinery/pkg/runtime"
	"k8s.io/apimachinery/pkg/util/validation/field"
	ctrl "sigs.k8s.io/controller-runtime"
	"sigs.k8s.io/controller-runtime/pkg/client"
	"sigs.k8s.io/controller-runtime/pkg/webhook"

	clusterv1 "sigs.k8s.io/cluster-api/api/v1beta1"
	"sigs.k8s.io/cluster-api/feature"
	"sigs.k8s.io/cluster-api/internal/topology/check"
	"sigs.k8s.io/cluster-api/internal/topology/variables"
	"sigs.k8s.io/cluster-api/util/version"
)

// SetupWebhookWithManager sets up Cluster webhooks.
func (webhook *Cluster) SetupWebhookWithManager(mgr ctrl.Manager) error {
	return ctrl.NewWebhookManagedBy(mgr).
		For(&clusterv1.Cluster{}).
		WithDefaulter(webhook).
		WithValidator(webhook).
		Complete()
}

// +kubebuilder:webhook:verbs=create;update;delete,path=/validate-cluster-x-k8s-io-v1beta1-cluster,mutating=false,failurePolicy=fail,matchPolicy=Equivalent,groups=cluster.x-k8s.io,resources=clusters,versions=v1beta1,name=validation.cluster.cluster.x-k8s.io,sideEffects=None,admissionReviewVersions=v1;v1beta1
// +kubebuilder:webhook:verbs=create;update,path=/mutate-cluster-x-k8s-io-v1beta1-cluster,mutating=true,failurePolicy=fail,matchPolicy=Equivalent,groups=cluster.x-k8s.io,resources=clusters,versions=v1beta1,name=default.cluster.cluster.x-k8s.io,sideEffects=None,admissionReviewVersions=v1;v1beta1

// Cluster implements a validating and defaulting webhook for Cluster.
type Cluster struct {
	Client client.Reader
}

var _ webhook.CustomDefaulter = &Cluster{}
var _ webhook.CustomValidator = &Cluster{}

// Default satisfies the defaulting webhook interface.
func (webhook *Cluster) Default(ctx context.Context, obj runtime.Object) error {
	cluster, ok := obj.(*clusterv1.Cluster)
	if !ok {
		return apierrors.NewBadRequest(fmt.Sprintf("expected a Cluster but got a %T", obj))
	}

	if cluster.Spec.InfrastructureRef != nil && cluster.Spec.InfrastructureRef.Namespace == "" {
		cluster.Spec.InfrastructureRef.Namespace = cluster.Namespace
	}

	if cluster.Spec.ControlPlaneRef != nil && cluster.Spec.ControlPlaneRef.Namespace == "" {
		cluster.Spec.ControlPlaneRef.Namespace = cluster.Namespace
	}

	// Additional defaulting if the Cluster uses a managed topology.
	if cluster.Spec.Topology != nil {
		// Tolerate version strings without a "v" prefix: prepend it if it's not there.
		if !strings.HasPrefix(cluster.Spec.Topology.Version, "v") {
			cluster.Spec.Topology.Version = "v" + cluster.Spec.Topology.Version
		}
		clusterClass, err := webhook.getClusterClassForCluster(ctx, cluster)
		if err != nil {
			// Return early with errors if the ClusterClass can't be retrieved.
			return apierrors.NewInternalError(errors.Wrapf(err, "Cluster %s can't be validated. ClusterClass %s can not be retrieved", cluster.Name, cluster.Spec.Topology.Class))
		}

		// We gather all defaulting errors and return them together.
		var allErrs field.ErrorList

		defaultedVariables, errs := variables.DefaultClusterVariables(cluster.Spec.Topology.Variables, clusterClass.Spec.Variables,
			field.NewPath("spec", "topology", "variables"))
		if len(errs) > 0 {
			allErrs = append(allErrs, errs...)
		} else {
			cluster.Spec.Topology.Variables = defaultedVariables
		}

		if cluster.Spec.Topology.Workers != nil {
			for i, md := range cluster.Spec.Topology.Workers.MachineDeployments {
				// Continue if there are no variable overrides.
				if md.Variables == nil || len(md.Variables.Overrides) == 0 {
					continue
				}

				defaultedVariables, errs := variables.DefaultMachineDeploymentVariables(md.Variables.Overrides, clusterClass.Spec.Variables,
					field.NewPath("spec", "topology", "workers", "machineDeployments").Index(i).Child("variables", "overrides"))
				if len(errs) > 0 {
					allErrs = append(allErrs, errs...)
				} else {
					md.Variables.Overrides = defaultedVariables
				}
			}
		}

		if len(allErrs) > 0 {
			return apierrors.NewInvalid(clusterv1.GroupVersion.WithKind("Cluster").GroupKind(), cluster.Name, allErrs)
		}
	}
	return nil
}

// ValidateCreate implements webhook.CustomValidator so a webhook will be registered for the type.
func (webhook *Cluster) ValidateCreate(ctx context.Context, obj runtime.Object) error {
	cluster, ok := obj.(*clusterv1.Cluster)
	if !ok {
		return apierrors.NewBadRequest(fmt.Sprintf("expected a Cluster but got a %T", obj))
	}
	return webhook.validate(ctx, nil, cluster)
}

// ValidateUpdate implements webhook.CustomValidator so a webhook will be registered for the type.
func (webhook *Cluster) ValidateUpdate(ctx context.Context, oldObj, newObj runtime.Object) error {
	newCluster, ok := newObj.(*clusterv1.Cluster)
	if !ok {
		return apierrors.NewBadRequest(fmt.Sprintf("expected a Cluster but got a %T", newObj))
	}
	oldCluster, ok := oldObj.(*clusterv1.Cluster)
	if !ok {
		return apierrors.NewBadRequest(fmt.Sprintf("expected a Cluster but got a %T", oldObj))
	}
	return webhook.validate(ctx, oldCluster, newCluster)
}

// ValidateDelete implements webhook.CustomValidator so a webhook will be registered for the type.
func (webhook *Cluster) ValidateDelete(_ context.Context, obj runtime.Object) error {
	return nil
}

func (webhook *Cluster) validate(ctx context.Context, oldCluster, newCluster *clusterv1.Cluster) error {
	var allErrs field.ErrorList
	specPath := field.NewPath("spec")
	if newCluster.Spec.InfrastructureRef != nil && newCluster.Spec.InfrastructureRef.Namespace != newCluster.Namespace {
		allErrs = append(
			allErrs,
			field.Invalid(
				specPath.Child("infrastructureRef", "namespace"),
				newCluster.Spec.InfrastructureRef.Namespace,
				"must match metadata.namespace",
			),
		)
	}

	if newCluster.Spec.ControlPlaneRef != nil && newCluster.Spec.ControlPlaneRef.Namespace != newCluster.Namespace {
		allErrs = append(
			allErrs,
			field.Invalid(
				specPath.Child("controlPlaneRef", "namespace"),
				newCluster.Spec.ControlPlaneRef.Namespace,
				"must match metadata.namespace",
			),
		)
	}
	topologyPath := specPath.Child("topology")

	// Validate the managed topology, if defined.
	if newCluster.Spec.Topology != nil {
		allErrs = append(allErrs, webhook.validateTopology(ctx, oldCluster, newCluster, topologyPath)...)
	}

	// On update.
	if oldCluster != nil {
		// Error if the update moves the cluster from Managed to Unmanaged i.e. the managed topology is removed on update.
		if oldCluster.Spec.Topology != nil && newCluster.Spec.Topology == nil {
			allErrs = append(allErrs, field.Forbidden(
				topologyPath,
				"cannot be removed from an existing Cluster",
			))
		}
	}

	if len(allErrs) > 0 {
		return apierrors.NewInvalid(clusterv1.GroupVersion.WithKind("Cluster").GroupKind(), newCluster.Name, allErrs)
	}
	return nil
}

func (webhook *Cluster) validateTopology(ctx context.Context, oldCluster, newCluster *clusterv1.Cluster, fldPath *field.Path) field.ErrorList {
	// NOTE: ClusterClass and managed topologies are behind ClusterTopology feature gate flag; the web hook
	// must prevent the usage of Cluster.Topology in case the feature flag is disabled.
	if !feature.Gates.Enabled(feature.ClusterTopology) {
		return field.ErrorList{
			field.Forbidden(
				fldPath,
				"can be set only if the ClusterTopology feature flag is enabled",
			),
		}
	}

	var allErrs field.ErrorList

	// class should be defined.
	if newCluster.Spec.Topology.Class == "" {
		allErrs = append(
			allErrs,
			field.Required(
				fldPath.Child("class"),
				"class cannot be empty",
			),
		)
	}

	// version should be valid.
	if !version.KubeSemver.MatchString(newCluster.Spec.Topology.Version) {
		allErrs = append(
			allErrs,
			field.Invalid(
				fldPath.Child("version"),
				newCluster.Spec.Topology.Version,
				"version must be a valid semantic version",
			),
		)
	}

	// clusterClass must exist.
	clusterClass := &clusterv1.ClusterClass{}
	// Check to see if the ClusterClass referenced in the Cluster currently exists.
	if err := webhook.Client.Get(ctx, client.ObjectKey{Namespace: newCluster.Namespace, Name: newCluster.Spec.Topology.Class}, clusterClass); err != nil {
		allErrs = append(
			allErrs, field.Invalid(
				fldPath.Child("class"),
				newCluster.Name,
				fmt.Sprintf("ClusterClass with name %q could not be found", newCluster.Spec.Topology.Class)))
		return allErrs
	}

	allErrs = append(allErrs, check.MachineDeploymentTopologiesAreValidAndDefinedInClusterClass(newCluster, clusterClass)...)

	// Check if the variables defined in the ClusterClass are valid.
	allErrs = append(allErrs, variables.ValidateClusterVariables(newCluster.Spec.Topology.Variables, clusterClass.Spec.Variables,
		fldPath.Child("variables"))...)

	// validate the MachineHealthChecks defined in the cluster topology
	allErrs = append(allErrs, validateMachineHealthChecks(newCluster, clusterClass)...)

	if newCluster.Spec.Topology.Workers != nil {
		for i, md := range newCluster.Spec.Topology.Workers.MachineDeployments {
			// Continue if there are no variable overrides.
			if md.Variables == nil || len(md.Variables.Overrides) == 0 {
				continue
			}

			allErrs = append(allErrs, variables.ValidateMachineDeploymentVariables(md.Variables.Overrides, clusterClass.Spec.Variables,
				fldPath.Child("workers", "machineDeployments").Index(i).Child("variables", "overrides"))...)
		}
	}

	if oldCluster != nil { // On update
		// Topology or Class can not be added on update unless unsafe cluster topology update annotation is set
		if oldCluster.Spec.Topology == nil || oldCluster.Spec.Topology.Class == "" {
			if _, ok := newCluster.Annotations[clusterv1.ClusterTopologyUnsafeUpdateClassNameAnnotation]; ok {
				return allErrs
			}

			allErrs = append(
				allErrs,
				field.Forbidden(
					fldPath.Child("class"),
					"class cannot be set on an existing Cluster",
				),
			)
			// return early here if there is no class to compare.
			return allErrs
		}

		// Version could only be increased.
		inVersion, err := semver.ParseTolerant(newCluster.Spec.Topology.Version)
		if err != nil {
			allErrs = append(
				allErrs,
				field.Invalid(
					fldPath.Child("version"),
					newCluster.Spec.Topology.Version,
					"version must be a valid semantic version",
				),
			)
		}
		oldVersion, err := semver.ParseTolerant(oldCluster.Spec.Topology.Version)
		if err != nil {
			// NOTE: this should never happen. Nevertheless, handling this for extra caution.
			allErrs = append(
				allErrs,
				field.Invalid(
					fldPath.Child("version"),
					oldCluster.Spec.Topology.Version,
					fmt.Sprintf("old version %q cannot be compared with %q", oldVersion, inVersion),
				),
			)
		}
		if inVersion.NE(semver.Version{}) && oldVersion.NE(semver.Version{}) && version.Compare(inVersion, oldVersion, version.WithBuildTags()) == -1 {
			allErrs = append(
				allErrs,
				field.Invalid(
					fldPath.Child("version"),
					newCluster.Spec.Topology.Version,
					fmt.Sprintf("version cannot be decreased from %q to %q", oldVersion, inVersion),
				),
			)
		}
		// A +2 minor version upgrade is not allowed.
		ceilVersion := semver.Version{
			Major: oldVersion.Major,
			Minor: oldVersion.Minor + 2,
			Patch: 0,
		}
		if inVersion.GTE(ceilVersion) {
			allErrs = append(
				allErrs,
				field.Forbidden(
					fldPath.Child("version"),
					fmt.Sprintf("version cannot be increased from %q to %q", oldVersion, inVersion),
				),
			)
		}

		// If the ClusterClass referenced in the Topology has changed compatibility checks are needed.
		if oldCluster.Spec.Topology.Class != newCluster.Spec.Topology.Class {
			// Check to see if the ClusterClass referenced in the old version of the Cluster exists.
			oldClusterClass, err := webhook.getClusterClassForCluster(ctx, oldCluster)
			if err != nil {
				allErrs = append(
					allErrs, field.Forbidden(
						fldPath.Child("class"),
						fmt.Sprintf("ClusterClass with name %q could not be found, change from class %[1]q to class %q cannot be validated",
							oldCluster.Spec.Topology.Class, newCluster.Spec.Topology.Class)))

				// Return early with errors if the ClusterClass can't be retrieved.
				return allErrs
			}

			// Check if the new and old ClusterClasses are compatible with one another.
			allErrs = append(allErrs, check.ClusterClassesAreCompatible(oldClusterClass, clusterClass)...)
		}
	}
	return allErrs
}

func (webhook *Cluster) getClusterClassForCluster(ctx context.Context, cluster *clusterv1.Cluster) (*clusterv1.ClusterClass, error) {
	clusterClass := &clusterv1.ClusterClass{}
	// Check to see if the ClusterClass referenced in the old version of the Cluster exists.
	if err := webhook.Client.Get(ctx, client.ObjectKey{Namespace: cluster.Namespace, Name: cluster.Spec.Topology.Class}, clusterClass); err != nil {
		return nil, err
	}
	return clusterClass, nil
}

func validateMachineHealthChecks(cluster *clusterv1.Cluster, clusterClass *clusterv1.ClusterClass) field.ErrorList {
	var allErrs field.ErrorList

	// Validate ControlPlane MachineHealthCheck if defined.
	if cluster.Spec.Topology.ControlPlane.MachineHealthCheck != nil && !cluster.Spec.Topology.ControlPlane.MachineHealthCheck.MachineHealthCheckClass.IsZero() {
		// Ensure ControlPlane does not define a MachineHealthCheck if the ClusterClass does not define MachineInfrastructure.
		if clusterClass.Spec.ControlPlane.MachineInfrastructure == nil {
			allErrs = append(allErrs, field.Forbidden(
				field.NewPath("spec", "topology", "controlPlane", "machineHealthCheck"),
				"can be set only if spec.controlPlane.machineInfrastructure is set in ClusterClass",
			))
		}
		// Ensure ControlPlane MachineHealthCheck defines UnhealthyConditions.
		if len(cluster.Spec.Topology.ControlPlane.MachineHealthCheck.MachineHealthCheckClass.UnhealthyConditions) == 0 {
			allErrs = append(allErrs, field.Forbidden(
				field.NewPath("spec", "topology", "controlPlane", "machineHealthCheck", "unhealthyConditions"),
				"must have at least one value",
			))
		}
	}

	// If MachineHealthCheck is explicitly enabled then make sure that a MachineHealthCheck definition is
	// available either in the Cluster topology or in the ClusterClass.
	// (One of these definitions will be used in the controller to create the MachineHealthCheck)
	if cluster.Spec.Topology.ControlPlane.MachineHealthCheck != nil &&
		cluster.Spec.Topology.ControlPlane.MachineHealthCheck.Enable != nil &&
		*cluster.Spec.Topology.ControlPlane.MachineHealthCheck.Enable &&
		cluster.Spec.Topology.ControlPlane.MachineHealthCheck.MachineHealthCheckClass.IsZero() &&
		clusterClass.Spec.ControlPlane.MachineHealthCheck == nil {
		allErrs = append(allErrs, field.Forbidden(
			field.NewPath("spec", "topology", "controlPlane", "machineHealthCheck", "enable"),
			fmt.Sprintf("cannot be set to %t as MachineHealthCheck definition is not available in the Cluster topology or the ClusterClass", *cluster.Spec.Topology.ControlPlane.MachineHealthCheck.Enable),
		))
	}

	if cluster.Spec.Topology.Workers != nil {
		for i, md := range cluster.Spec.Topology.Workers.MachineDeployments {
			// If MachineHealthCheck is defined ensure it defines UnhealthyConditions.
			if md.MachineHealthCheck != nil && !md.MachineHealthCheck.MachineHealthCheckClass.IsZero() {
				if len(md.MachineHealthCheck.MachineHealthCheckClass.UnhealthyConditions) == 0 {
					allErrs = append(allErrs, field.Forbidden(
						field.NewPath("spec", "topology", "workers", "machineDeployments", "machineHealthCheck").Index(i).Child("unhealthyConditions"),
						"must have at least one value",
					))
				}
			}

			// If MachineHealthCheck is explicitly enabled then make sure that a MachineHealthCheck definition is
			// available either in the Cluster topology or in the ClusterClass.
			// (One of these definitions will be used in the controller to create the MachineHealthCheck)
			mdClass := machineDeploymentClassOfName(clusterClass, md.Class)
			if mdClass != nil { // Note: we skip handling the nil case here as it is already handled in previous validations.
				if md.MachineHealthCheck != nil &&
					md.MachineHealthCheck.Enable != nil &&
					*md.MachineHealthCheck.Enable &&
					md.MachineHealthCheck.MachineHealthCheckClass.IsZero() &&
					mdClass.MachineHealthCheck == nil {
					allErrs = append(allErrs, field.Forbidden(
						field.NewPath("spec", "topology", "workers", "machineDeployments", "machineHealthCheck").Index(i).Child("enable"),
						fmt.Sprintf("cannot be set to %t as MachineHealthCheck definition is not available in the Cluster topology or the ClusterClass", *md.MachineHealthCheck.Enable),
					))
				}
			}
		}
	}

	return allErrs
}

// machineDeploymentClassOfName find a MachineDeploymentClass of the given name in the provided ClusterClass.
// Returns nill if can not find one.
// TODO: Check if there is already a helper function that can do this.
func machineDeploymentClassOfName(clusterClass *clusterv1.ClusterClass, name string) *clusterv1.MachineDeploymentClass {
	for _, mdClass := range clusterClass.Spec.Workers.MachineDeployments {
		if mdClass.Class == name {
			return &mdClass
		}
	}
	return nil
}