velero/pkg/controller/pod_volume_backup_controller.go

/*
Copyright the Velero contributors.

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 controller

import (
	"context"
	"encoding/json"
	"fmt"
	"os"
	"path/filepath"
	"strings"
	"time"

	jsonpatch "github.com/evanphx/json-patch"
	"github.com/pkg/errors"
	"github.com/sirupsen/logrus"
	apierrors "k8s.io/apimachinery/pkg/api/errors"
	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
	"k8s.io/apimachinery/pkg/labels"
	"k8s.io/apimachinery/pkg/types"
	"k8s.io/apimachinery/pkg/util/clock"
	corev1informers "k8s.io/client-go/informers/core/v1"
	corev1listers "k8s.io/client-go/listers/core/v1"
	"k8s.io/client-go/tools/cache"

	velerov1api "github.com/vmware-tanzu/velero/pkg/apis/velero/v1"
	velerov1client "github.com/vmware-tanzu/velero/pkg/generated/clientset/versioned/typed/velero/v1"
	informers "github.com/vmware-tanzu/velero/pkg/generated/informers/externalversions/velero/v1"
	listers "github.com/vmware-tanzu/velero/pkg/generated/listers/velero/v1"
	"github.com/vmware-tanzu/velero/pkg/metrics"
	"github.com/vmware-tanzu/velero/pkg/restic"
	"github.com/vmware-tanzu/velero/pkg/util/filesystem"
	"github.com/vmware-tanzu/velero/pkg/util/kube"

	"sigs.k8s.io/controller-runtime/pkg/client"
)

type podVolumeBackupController struct {
	*genericController

	podVolumeBackupClient velerov1client.PodVolumeBackupsGetter
	podVolumeBackupLister listers.PodVolumeBackupLister
	podLister             corev1listers.PodLister
	pvcLister             corev1listers.PersistentVolumeClaimLister
	pvLister              corev1listers.PersistentVolumeLister
	kbClient              client.Client
	nodeName              string
	metrics               *metrics.ServerMetrics

	processBackupFunc func(*velerov1api.PodVolumeBackup) error
	fileSystem        filesystem.Interface
	clock             clock.Clock
}

// NewPodVolumeBackupController creates a new pod volume backup controller.
func NewPodVolumeBackupController(
	logger logrus.FieldLogger,
	podVolumeBackupInformer informers.PodVolumeBackupInformer,
	podVolumeBackupClient velerov1client.PodVolumeBackupsGetter,
	podInformer cache.SharedIndexInformer,
	pvcInformer corev1informers.PersistentVolumeClaimInformer,
	pvInformer corev1informers.PersistentVolumeInformer,
	metrics *metrics.ServerMetrics,
	kbClient client.Client,
	nodeName string,
) Interface {
	c := &podVolumeBackupController{
		genericController:     newGenericController(PodVolumeBackup, logger),
		podVolumeBackupClient: podVolumeBackupClient,
		podVolumeBackupLister: podVolumeBackupInformer.Lister(),
		podLister:             corev1listers.NewPodLister(podInformer.GetIndexer()),
		pvcLister:             pvcInformer.Lister(),
		pvLister:              pvInformer.Lister(),
		kbClient:              kbClient,
		nodeName:              nodeName,
		metrics:               metrics,

		fileSystem: filesystem.NewFileSystem(),
		clock:      &clock.RealClock{},
	}

	c.syncHandler = c.processQueueItem
	c.cacheSyncWaiters = append(
		c.cacheSyncWaiters,
		podVolumeBackupInformer.Informer().HasSynced,
		podInformer.HasSynced,
		pvcInformer.Informer().HasSynced,
	)
	c.processBackupFunc = c.processBackup

	podVolumeBackupInformer.Informer().AddEventHandler(
		cache.ResourceEventHandlerFuncs{
			AddFunc:    c.pvbHandler,
			UpdateFunc: func(_, obj interface{}) { c.pvbHandler(obj) },
		},
	)

	return c
}

func (c *podVolumeBackupController) pvbHandler(obj interface{}) {
	req := obj.(*velerov1api.PodVolumeBackup)

	// only enqueue items for this node
	if req.Spec.Node != c.nodeName {
		return
	}

	log := loggerForPodVolumeBackup(c.logger, req)

	if req.Status.Phase != "" && req.Status.Phase != velerov1api.PodVolumeBackupPhaseNew {
		log.Debug("Backup is not new, not enqueuing")
		return
	}

	c.metrics.RegisterPodVolumeBackupEnqueue(c.nodeName)

	log.Debug("Enqueueing")
	c.enqueue(obj)
}

func (c *podVolumeBackupController) processQueueItem(key string) error {
	log := c.logger.WithField("key", key)
	log.Debug("Running processQueueItem")

	ns, name, err := cache.SplitMetaNamespaceKey(key)
	if err != nil {
		log.WithError(err).Error("error splitting queue key")
		return nil
	}

	req, err := c.podVolumeBackupLister.PodVolumeBackups(ns).Get(name)
	if apierrors.IsNotFound(err) {
		log.Debug("Unable to find PodVolumeBackup")
		return nil
	}
	if err != nil {
		return errors.Wrap(err, "error getting PodVolumeBackup")
	}

	// only process new items
	switch req.Status.Phase {
	case "", velerov1api.PodVolumeBackupPhaseNew:
	default:
		return nil
	}

	// Don't mutate the shared cache
	reqCopy := req.DeepCopy()
	return c.processBackupFunc(reqCopy)
}

func loggerForPodVolumeBackup(baseLogger logrus.FieldLogger, req *velerov1api.PodVolumeBackup) logrus.FieldLogger {
	log := baseLogger.WithFields(logrus.Fields{
		"namespace": req.Namespace,
		"name":      req.Name,
	})

	if len(req.OwnerReferences) == 1 {
		log = log.WithField("backup", fmt.Sprintf("%s/%s", req.Namespace, req.OwnerReferences[0].Name))
	}

	return log
}

func getOwningBackup(req *velerov1api.PodVolumeBackup) string {
	return fmt.Sprintf("%s/%s", req.Namespace, req.OwnerReferences[0].Name)
}

func (c *podVolumeBackupController) processBackup(req *velerov1api.PodVolumeBackup) error {
	log := loggerForPodVolumeBackup(c.logger, req)

	log.Info("Backup starting")

	var err error

	// update status to InProgress
	req, err = c.patchPodVolumeBackup(req, func(r *velerov1api.PodVolumeBackup) {
		r.Status.Phase = velerov1api.PodVolumeBackupPhaseInProgress
		r.Status.StartTimestamp = &metav1.Time{Time: c.clock.Now()}
	})
	if err != nil {
		log.WithError(err).Error("Error setting PodVolumeBackup StartTimestamp and phase to InProgress")
		return errors.WithStack(err)
	}

	pod, err := c.podLister.Pods(req.Spec.Pod.Namespace).Get(req.Spec.Pod.Name)
	if err != nil {
		log.WithError(err).Errorf("Error getting pod %s/%s", req.Spec.Pod.Namespace, req.Spec.Pod.Name)
		return c.fail(req, errors.Wrap(err, "error getting pod").Error(), log)
	}

	volumeDir, err := kube.GetVolumeDirectory(pod, req.Spec.Volume, c.pvcLister, c.pvLister)
	if err != nil {
		log.WithError(err).Error("Error getting volume directory name")
		return c.fail(req, errors.Wrap(err, "error getting volume directory name").Error(), log)
	}

	pathGlob := fmt.Sprintf("/host_pods/%s/volumes/*/%s", string(req.Spec.Pod.UID), volumeDir)
	log.WithField("pathGlob", pathGlob).Debug("Looking for path matching glob")

	path, err := singlePathMatch(pathGlob)
	if err != nil {
		log.WithError(err).Error("Error uniquely identifying volume path")
		return c.fail(req, errors.Wrap(err, "error getting volume path on host").Error(), log)
	}
	log.WithField("path", path).Debugf("Found path matching glob")

	// temp creds
	credentialsFile, err := restic.TempCredentialsFile(c.kbClient, req.Namespace, req.Spec.Pod.Namespace, c.fileSystem)
	if err != nil {
		log.WithError(err).Error("Error creating temp restic credentials file")
		return c.fail(req, errors.Wrap(err, "error creating temp restic credentials file").Error(), log)
	}
	// ignore error since there's nothing we can do and it's a temp file.
	defer os.Remove(credentialsFile)

	resticCmd := restic.BackupCommand(
		req.Spec.RepoIdentifier,
		credentialsFile,
		path,
		req.Spec.Tags,
	)

	// if there's a caCert on the ObjectStorage, write it to disk so that it can be passed to restic
	caCert, err := restic.GetCACert(c.kbClient, req.Namespace, req.Spec.BackupStorageLocation)
	if err != nil {
		log.WithError(err).Error("Error getting caCert")
	}

	var caCertFile string
	if caCert != nil {
		caCertFile, err = restic.TempCACertFile(caCert, req.Spec.BackupStorageLocation, c.fileSystem)
		if err != nil {
			log.WithError(err).Error("Error creating temp cacert file")
		}
		// ignore error since there's nothing we can do and it's a temp file.
		defer os.Remove(caCertFile)
	}
	resticCmd.CACertFile = caCertFile

	// Running restic command might need additional provider specific environment variables. Based on the provider, we
	// set resticCmd.Env appropriately (currently for Azure and S3 based backuplocations)
	var env []string
	if strings.HasPrefix(req.Spec.RepoIdentifier, "azure") {
		if env, err = restic.AzureCmdEnv(c.kbClient, req.Namespace, req.Spec.BackupStorageLocation); err != nil {
			return c.fail(req, errors.Wrap(err, "error setting restic cmd env").Error(), log)
		}
		resticCmd.Env = env
	} else if strings.HasPrefix(req.Spec.RepoIdentifier, "s3") {
		if env, err = restic.S3CmdEnv(c.kbClient, req.Namespace, req.Spec.BackupStorageLocation); err != nil {
			return c.fail(req, errors.Wrap(err, "error setting restic cmd env").Error(), log)
		}
		resticCmd.Env = env
	}

	// If this is a PVC, look for the most recent completed pod volume backup for it and get
	// its restic snapshot ID to use as the value of the `--parent` flag. Without this,
	// if the pod using the PVC (and therefore the directory path under /host_pods/) has
	// changed since the PVC's last backup, restic will not be able to identify a suitable
	// parent snapshot to use, and will have to do a full rescan of the contents of the PVC.
	if pvcUID, ok := req.Labels[velerov1api.PVCUIDLabel]; ok {
		parentSnapshotID := getParentSnapshot(log, pvcUID, req.Spec.BackupStorageLocation, c.podVolumeBackupLister.PodVolumeBackups(req.Namespace))
		if parentSnapshotID == "" {
			log.Info("No parent snapshot found for PVC, not using --parent flag for this backup")
		} else {
			log.WithField("parentSnapshotID", parentSnapshotID).Info("Setting --parent flag for this backup")
			resticCmd.ExtraFlags = append(resticCmd.ExtraFlags, fmt.Sprintf("--parent=%s", parentSnapshotID))
		}
	}

	var stdout, stderr string

	var emptySnapshot bool
	if stdout, stderr, err = restic.RunBackup(resticCmd, log, c.updateBackupProgressFunc(req, log)); err != nil {
		if strings.Contains(stderr, "snapshot is empty") {
			emptySnapshot = true
		} else {
			log.WithError(errors.WithStack(err)).Errorf("Error running command=%s, stdout=%s, stderr=%s", resticCmd.String(), stdout, stderr)
			return c.fail(req, fmt.Sprintf("error running restic backup, stderr=%s: %s", stderr, err.Error()), log)
		}
	}
	log.Debugf("Ran command=%s, stdout=%s, stderr=%s", resticCmd.String(), stdout, stderr)

	var snapshotID string
	if !emptySnapshot {
		snapshotID, err = restic.GetSnapshotID(req.Spec.RepoIdentifier, credentialsFile, req.Spec.Tags, env, caCertFile)
		if err != nil {
			log.WithError(err).Error("Error getting SnapshotID")
			return c.fail(req, errors.Wrap(err, "error getting snapshot id").Error(), log)
		}
	}

	// update status to Completed with path & snapshot id
	req, err = c.patchPodVolumeBackup(req, func(r *velerov1api.PodVolumeBackup) {
		r.Status.Path = path
		r.Status.Phase = velerov1api.PodVolumeBackupPhaseCompleted
		r.Status.SnapshotID = snapshotID
		r.Status.CompletionTimestamp = &metav1.Time{Time: c.clock.Now()}
		if emptySnapshot {
			r.Status.Message = "volume was empty so no snapshot was taken"
		}
	})
	if err != nil {
		log.WithError(err).Error("Error setting PodVolumeBackup phase to Completed")
		return err
	}
	latencyDuration := req.Status.CompletionTimestamp.Time.Sub(req.Status.StartTimestamp.Time)
	latencySeconds := float64(latencyDuration / time.Second)
	backupName := getOwningBackup(req)
	c.metrics.ObserveResticOpLatency(c.nodeName, req.Name, resticCmd.Command, backupName, latencySeconds)
	c.metrics.RegisterResticOpLatencyGauge(c.nodeName, req.Name, resticCmd.Command, backupName, latencySeconds)
	c.metrics.RegisterPodVolumeBackupDequeue(c.nodeName)
	log.Info("Backup completed")

	return nil
}

// getParentSnapshot finds the most recent completed pod volume backup for the specified PVC and returns its
// restic snapshot ID. Any errors encountered are logged but not returned since they do not prevent a backup
// from proceeding.
func getParentSnapshot(log logrus.FieldLogger, pvcUID, backupStorageLocation string, podVolumeBackupLister listers.PodVolumeBackupNamespaceLister) string {
	log = log.WithField("pvcUID", pvcUID)
	log.Infof("Looking for most recent completed pod volume backup for this PVC")

	pvcBackups, err := podVolumeBackupLister.List(labels.SelectorFromSet(map[string]string{velerov1api.PVCUIDLabel: pvcUID}))
	if err != nil {
		log.WithError(errors.WithStack(err)).Error("Error listing pod volume backups for PVC")
		return ""
	}

	// go through all the pod volume backups for the PVC and look for the most recent completed one
	// to use as the parent.
	var mostRecentBackup *velerov1api.PodVolumeBackup
	for _, backup := range pvcBackups {
		if backup.Status.Phase != velerov1api.PodVolumeBackupPhaseCompleted {
			continue
		}

		if backupStorageLocation != backup.Spec.BackupStorageLocation {
			// Check the backup storage location is the same as spec in order to support backup to multiple backup-locations.
			// Otherwise, there exists a case that backup volume snapshot to the second location would failed, since the founded
			// parent ID is only valid for the first backup location, not the second backup location.
			// Also, the second backup should not use the first backup parent ID since its for the first backup location only.
			continue
		}

		if mostRecentBackup == nil || backup.Status.StartTimestamp.After(mostRecentBackup.Status.StartTimestamp.Time) {
			mostRecentBackup = backup
		}
	}

	if mostRecentBackup == nil {
		log.Info("No completed pod volume backup found for PVC")
		return ""
	}

	log.WithFields(map[string]interface{}{
		"parentPodVolumeBackup": mostRecentBackup.Name,
		"parentSnapshotID":      mostRecentBackup.Status.SnapshotID,
	}).Info("Found most recent completed pod volume backup for PVC")

	return mostRecentBackup.Status.SnapshotID
}

func (c *podVolumeBackupController) patchPodVolumeBackup(req *velerov1api.PodVolumeBackup, mutate func(*velerov1api.PodVolumeBackup)) (*velerov1api.PodVolumeBackup, error) {
	// Record original json
	oldData, err := json.Marshal(req)
	if err != nil {
		return nil, errors.Wrap(err, "error marshalling original PodVolumeBackup")
	}

	// Mutate
	mutate(req)

	// Record new json
	newData, err := json.Marshal(req)
	if err != nil {
		return nil, errors.Wrap(err, "error marshalling updated PodVolumeBackup")
	}

	patchBytes, err := jsonpatch.CreateMergePatch(oldData, newData)
	if err != nil {
		return nil, errors.Wrap(err, "error creating json merge patch for PodVolumeBackup")
	}

	req, err = c.podVolumeBackupClient.PodVolumeBackups(req.Namespace).Patch(context.TODO(), req.Name, types.MergePatchType, patchBytes, metav1.PatchOptions{})
	if err != nil {
		return nil, errors.Wrap(err, "error patching PodVolumeBackup")
	}

	return req, nil
}

// updateBackupProgressFunc returns a func that takes progress info and patches
// the PVB with the new progress
func (c *podVolumeBackupController) updateBackupProgressFunc(req *velerov1api.PodVolumeBackup, log logrus.FieldLogger) func(velerov1api.PodVolumeOperationProgress) {
	return func(progress velerov1api.PodVolumeOperationProgress) {
		if _, err := c.patchPodVolumeBackup(req, func(r *velerov1api.PodVolumeBackup) {
			r.Status.Progress = progress
		}); err != nil {
			log.WithError(err).Error("error updating PodVolumeBackup progress")
		}
	}
}

func (c *podVolumeBackupController) fail(req *velerov1api.PodVolumeBackup, msg string, log logrus.FieldLogger) error {
	if _, err := c.patchPodVolumeBackup(req, func(r *velerov1api.PodVolumeBackup) {
		r.Status.Phase = velerov1api.PodVolumeBackupPhaseFailed
		r.Status.Message = msg
		r.Status.CompletionTimestamp = &metav1.Time{Time: c.clock.Now()}
	}); err != nil {
		log.WithError(err).Error("Error setting PodVolumeBackup phase to Failed")
		return err
	}
	return nil
}

func singlePathMatch(path string) (string, error) {
	matches, err := filepath.Glob(path)
	if err != nil {
		return "", errors.WithStack(err)
	}

	if len(matches) != 1 {
		return "", errors.Errorf("expected one matching path, got %d", len(matches))
	}

	return matches[0], nil
}