md_remediations.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 e2e

import (
	"context"
	"fmt"
	"os"
	"path/filepath"

	. "github.com/onsi/ginkgo/v2"
	. "github.com/onsi/gomega"
	corev1 "k8s.io/api/core/v1"
	"k8s.io/utils/pointer"

	"sigs.k8s.io/cluster-api/test/framework"
	"sigs.k8s.io/cluster-api/test/framework/clusterctl"
	"sigs.k8s.io/cluster-api/util"
)

// MachineDeploymentRemediationSpecInput is the input for MachineDeploymentRemediationSpec.
type MachineDeploymentRemediationSpecInput struct {
	E2EConfig             *clusterctl.E2EConfig
	ClusterctlConfigPath  string
	BootstrapClusterProxy framework.ClusterProxy
	ArtifactFolder        string
	SkipCleanup           bool
	ControlPlaneWaiters   clusterctl.ControlPlaneWaiters

	// Flavor, if specified, must refer to a template that has a MachineHealthCheck
	// resource configured to match the MachineDeployment managed Machines and be
	// configured to treat "e2e.remediation.condition" "False" as an unhealthy
	// condition with a short timeout.
	// If not specified, "md-remediation" is used.
	Flavor *string
}

// MachineDeploymentRemediationSpec implements a test that verifies that Machines are remediated by MHC during unhealthy conditions.
func MachineDeploymentRemediationSpec(ctx context.Context, inputGetter func() MachineDeploymentRemediationSpecInput) {
	var (
		specName         = "md-remediation"
		input            MachineDeploymentRemediationSpecInput
		namespace        *corev1.Namespace
		cancelWatches    context.CancelFunc
		clusterResources *clusterctl.ApplyClusterTemplateAndWaitResult
	)

	BeforeEach(func() {
		Expect(ctx).NotTo(BeNil(), "ctx is required for %s spec", specName)
		input = inputGetter()
		Expect(input.E2EConfig).ToNot(BeNil(), "Invalid argument. input.E2EConfig can't be nil when calling %s spec", specName)
		Expect(input.ClusterctlConfigPath).To(BeAnExistingFile(), "Invalid argument. input.ClusterctlConfigPath must be an existing file when calling %s spec", specName)
		Expect(input.BootstrapClusterProxy).ToNot(BeNil(), "Invalid argument. input.BootstrapClusterProxy can't be nil when calling %s spec", specName)
		Expect(os.MkdirAll(input.ArtifactFolder, 0750)).To(Succeed(), "Invalid argument. input.ArtifactFolder can't be created for %s spec", specName)
		Expect(input.E2EConfig.Variables).To(HaveKey(KubernetesVersion))

		// Setup a Namespace where to host objects for this spec and create a watcher for the namespace events.
		namespace, cancelWatches = setupSpecNamespace(ctx, specName, input.BootstrapClusterProxy, input.ArtifactFolder)
		clusterResources = new(clusterctl.ApplyClusterTemplateAndWaitResult)
	})

	It("Should replace unhealthy machines", func() {
		By("Creating a workload cluster")

		clusterctl.ApplyClusterTemplateAndWait(ctx, clusterctl.ApplyClusterTemplateAndWaitInput{
			ClusterProxy: input.BootstrapClusterProxy,
			ConfigCluster: clusterctl.ConfigClusterInput{
				LogFolder:                filepath.Join(input.ArtifactFolder, "clusters", input.BootstrapClusterProxy.GetName()),
				ClusterctlConfigPath:     input.ClusterctlConfigPath,
				KubeconfigPath:           input.BootstrapClusterProxy.GetKubeconfigPath(),
				InfrastructureProvider:   clusterctl.DefaultInfrastructureProvider,
				Flavor:                   pointer.StringDeref(input.Flavor, "md-remediation"),
				Namespace:                namespace.Name,
				ClusterName:              fmt.Sprintf("%s-%s", specName, util.RandomString(6)),
				KubernetesVersion:        input.E2EConfig.GetVariable(KubernetesVersion),
				ControlPlaneMachineCount: pointer.Int64(1),
				WorkerMachineCount:       pointer.Int64(1),
			},
			ControlPlaneWaiters:          input.ControlPlaneWaiters,
			WaitForClusterIntervals:      input.E2EConfig.GetIntervals(specName, "wait-cluster"),
			WaitForControlPlaneIntervals: input.E2EConfig.GetIntervals(specName, "wait-control-plane"),
			WaitForMachineDeployments:    input.E2EConfig.GetIntervals(specName, "wait-worker-nodes"),
		}, clusterResources)

		// TODO: this should be re-written like the KCP remediation test, because the current implementation
		//   only tests that MHC applies the unhealthy condition but it doesn't test that the unhealthy machine is delete and a replacement machine comes up.
		By("Setting a machine unhealthy and wait for MachineDeployment remediation")
		framework.DiscoverMachineHealthChecksAndWaitForRemediation(ctx, framework.DiscoverMachineHealthCheckAndWaitForRemediationInput{
			ClusterProxy:              input.BootstrapClusterProxy,
			Cluster:                   clusterResources.Cluster,
			WaitForMachineRemediation: input.E2EConfig.GetIntervals(specName, "wait-machine-remediation"),
		})

		By("PASSED!")
	})

	AfterEach(func() {
		// Dumps all the resources in the spec namespace, then cleanups the cluster object and the spec namespace itself.
		dumpSpecResourcesAndCleanup(ctx, specName, input.BootstrapClusterProxy, input.ArtifactFolder, namespace, cancelWatches, clusterResources.Cluster, input.E2EConfig.GetIntervals, input.SkipCleanup)
	})
}