[Feature] Ensure consistency during cancellation of replication (#1134)

CHANGELOG.md

@@ -2,6 +2,7 @@
 
 ## [master](https://github.com/arangodb/kube-arangodb/tree/master) (N/A)
 - (Feature) Add action progress
+- (Feature) Ensure consistency during replication cancellation
 
 ## [1.2.19](https://github.com/arangodb/kube-arangodb/tree/1.2.19) (2022-10-05)
 - (Bugfix) Prevent changes when UID is wrong

pkg/apis/replication/v1/conditions.go

@@ -31,6 +31,17 @@ type ConditionType string
 const (
 	// ConditionTypeConfigured indicates that the replication has been configured.
 	ConditionTypeConfigured ConditionType = "Configured"
+	// ConditionTypeEnsuredInSync indicates that the replication consistency was checked.
+	ConditionTypeEnsuredInSync ConditionType = "EnsuredInSync"
+	// ConditionTypeAborted indicates that the replication was canceled with abort option.
+	ConditionTypeAborted ConditionType = "Aborted"
+
+	// ConditionConfiguredReasonActive describes synchronization as active.
+	ConditionConfiguredReasonActive = "Active"
+	// ConditionConfiguredReasonInactive describes synchronization as inactive.
+	ConditionConfiguredReasonInactive = "Inactive"
+	// ConditionConfiguredReasonInvalid describes synchronization as active.
+	ConditionConfiguredReasonInvalid = "Invalid"
 )
 
 // Condition represents one current condition of a deployment or deployment member.

pkg/apis/replication/v1/replication_spec.go

@@ -27,6 +27,18 @@ import "github.com/arangodb/kube-arangodb/pkg/util/errors"
 type DeploymentReplicationSpec struct {
 	Source      EndpointSpec `json:"source"`
 	Destination EndpointSpec `json:"destination"`
+	// Cancellation describes what to do during cancellation process.
+	Cancellation DeploymentReplicationCancel `json:"cancellation"`
+}
+
+// DeploymentReplicationCancel describes what to do during cancellation process.
+type DeploymentReplicationCancel struct {
+	// EnsureInSync if it is true then during cancellation process data consistency is required.
+	// Default value is true.
+	EnsureInSync *bool `json:"ensureInSync"`
+	// SourceReadOnly if it true then after cancellation source data center should be in read-only mode.
+	// Default value is false.
+	SourceReadOnly *bool `json:"sourceReadOnly"`
 }
 
 // Validate the given spec, returning an error on validation

pkg/apis/replication/v1/replication_status.go

@@ -36,6 +36,7 @@ type DeploymentReplicationStatus struct {
 	// Destination contains the detailed status of the destination endpoint
 	Destination EndpointStatus `json:"destination"`
 
+	// Deprecated: this field will not be updated anymore
 	// CancelFailures records the number of times that the configuration was canceled
 	// which resulted in an error.
 	CancelFailures int `json:"cancel-failures,omitempty"`

pkg/apis/replication/v1/zz_generated.deepcopy.go

@@ -211,11 +211,38 @@ func (in *DatabaseSynchronizationStatus) DeepCopy() *DatabaseSynchronizationStat
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *DeploymentReplicationCancel) DeepCopyInto(out *DeploymentReplicationCancel) {
+	*out = *in
+	if in.EnsureInSync != nil {
+		in, out := &in.EnsureInSync, &out.EnsureInSync
+		*out = new(bool)
+		**out = **in
+	}
+	if in.SourceReadOnly != nil {
+		in, out := &in.SourceReadOnly, &out.SourceReadOnly
+		*out = new(bool)
+		**out = **in
+	}
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeploymentReplicationCancel.
+func (in *DeploymentReplicationCancel) DeepCopy() *DeploymentReplicationCancel {
+	if in == nil {
+		return nil
+	}
+	out := new(DeploymentReplicationCancel)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *DeploymentReplicationSpec) DeepCopyInto(out *DeploymentReplicationSpec) {
 	*out = *in
 	in.Source.DeepCopyInto(&out.Source)
 	in.Destination.DeepCopyInto(&out.Destination)
+	in.Cancellation.DeepCopyInto(&out.Cancellation)
 	return
 }
 

pkg/apis/replication/v2alpha1/conditions.go

@@ -31,6 +31,17 @@ type ConditionType string
 const (
 	// ConditionTypeConfigured indicates that the replication has been configured.
 	ConditionTypeConfigured ConditionType = "Configured"
+	// ConditionTypeEnsuredInSync indicates that the replication consistency was checked.
+	ConditionTypeEnsuredInSync ConditionType = "EnsuredInSync"
+	// ConditionTypeAborted indicates that the replication is cancelling with abort option.
+	ConditionTypeAborted ConditionType = "Aborted"
+
+	// ConditionConfiguredReasonActive describes synchronization as active.
+	ConditionConfiguredReasonActive = "Active"
+	// ConditionConfiguredReasonInactive describes synchronization as inactive.
+	ConditionConfiguredReasonInactive = "Inactive"
+	// ConditionConfiguredReasonInvalid describes synchronization as active.
+	ConditionConfiguredReasonInvalid = "Invalid"
 )
 
 // Condition represents one current condition of a deployment or deployment member.

pkg/apis/replication/v2alpha1/replication_spec.go

@@ -27,6 +27,18 @@ import "github.com/arangodb/kube-arangodb/pkg/util/errors"
 type DeploymentReplicationSpec struct {
 	Source      EndpointSpec `json:"source"`
 	Destination EndpointSpec `json:"destination"`
+	// Cancellation describes what to do during cancellation process.
+	Cancellation DeploymentReplicationCancel `json:"cancellation"`
+}
+
+// DeploymentReplicationCancel describes what to do during cancellation process.
+type DeploymentReplicationCancel struct {
+	// EnsureInSync if it is true then during cancellation process data consistency is required.
+	// Default value is true.
+	EnsureInSync *bool `json:"ensureInSync"`
+	// SourceReadOnly if it true then after cancellation source data center should be in read-only mode.
+	// Default value is false.
+	SourceReadOnly *bool `json:"sourceReadOnly"`
 }
 
 // Validate the given spec, returning an error on validation

pkg/apis/replication/v2alpha1/replication_status.go

@@ -36,6 +36,7 @@ type DeploymentReplicationStatus struct {
 	// Destination contains the detailed status of the destination endpoint
 	Destination EndpointStatus `json:"destination"`
 
+	// Deprecated: this field will not be updated anymore
 	// CancelFailures records the number of times that the configuration was canceled
 	// which resulted in an error.
 	CancelFailures int `json:"cancel-failures,omitempty"`

pkg/apis/replication/v2alpha1/zz_generated.deepcopy.go

@@ -211,11 +211,38 @@ func (in *DatabaseSynchronizationStatus) DeepCopy() *DatabaseSynchronizationStat
 	return out
 }
 
+// DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
+func (in *DeploymentReplicationCancel) DeepCopyInto(out *DeploymentReplicationCancel) {
+	*out = *in
+	if in.EnsureInSync != nil {
+		in, out := &in.EnsureInSync, &out.EnsureInSync
+		*out = new(bool)
+		**out = **in
+	}
+	if in.SourceReadOnly != nil {
+		in, out := &in.SourceReadOnly, &out.SourceReadOnly
+		*out = new(bool)
+		**out = **in
+	}
+	return
+}
+
+// DeepCopy is an autogenerated deepcopy function, copying the receiver, creating a new DeploymentReplicationCancel.
+func (in *DeploymentReplicationCancel) DeepCopy() *DeploymentReplicationCancel {
+	if in == nil {
+		return nil
+	}
+	out := new(DeploymentReplicationCancel)
+	in.DeepCopyInto(out)
+	return out
+}
+
 // DeepCopyInto is an autogenerated deepcopy function, copying the receiver, writing into out. in must be non-nil.
 func (in *DeploymentReplicationSpec) DeepCopyInto(out *DeploymentReplicationSpec) {
 	*out = *in
 	in.Source.DeepCopyInto(&out.Source)
 	in.Destination.DeepCopyInto(&out.Destination)
+	in.Cancellation.DeepCopyInto(&out.Cancellation)
 	return
 }
 

pkg/replication/deployment_replication.go

@@ -22,6 +22,7 @@ package replication
 
 import (
 	"context"
+	"fmt"
 	"reflect"
 	"sync/atomic"
 	"time"
@@ -72,11 +73,13 @@ const (
 	deploymentReplicationEventQueueSize = 100
 	minInspectionInterval               = time.Second // Ensure we inspect the generated resources no less than with this interval
 	maxInspectionInterval               = time.Minute // Ensure we inspect the generated resources no less than with this interval
+	cancellationInterval                = time.Second * 5
 )
 
 // DeploymentReplication is the in process state of an ArangoDeploymentReplication.
 type DeploymentReplication struct {
 	log       logging.Logger
+	lastLog   time.Time
 	apiObject *api.ArangoDeploymentReplication // API object
 	status    api.DeploymentReplicationStatus  // Internal status of the CR
 	config    Config
@@ -246,8 +249,14 @@ func (dr *DeploymentReplication) createEvent(evt *k8sutil.Event) {
 	dr.deps.EventRecorder.Event(evt.InvolvedObject, evt.Type, evt.Reason, evt.Message)
 }
 
-// Update the status of the API object from the internal status
+// Update the status of the API object from the internal status.
+// Has no effect if object is being deleted.
 func (dr *DeploymentReplication) updateCRStatus() error {
+	if dr.apiObject.DeletionTimestamp != nil {
+		// Object is being removed so nothing can be changed in the resource.
+		// The field DeploymentReplication.status is updated automatically here.
+		return nil
+	}
 	if reflect.DeepEqual(dr.apiObject.Status, dr.status) {
 		// Nothing has changed
 		return nil
@@ -323,8 +332,13 @@ func (dr *DeploymentReplication) updateCRSpec(newSpec api.DeploymentReplicationS
 
 // failOnError reports the given error and sets the deployment replication status to failed.
 func (dr *DeploymentReplication) failOnError(err error, msg string) {
-	dr.log.Err(err).Error(msg)
-	dr.status.Reason = err.Error()
+	if err != nil {
+		dr.log.Err(err).Error(msg)
+		dr.status.Reason = fmt.Sprintf("%s: %s", msg, err.Error())
+	} else {
+		dr.log.Error(msg)
+		dr.status.Reason = msg
+	}
 	dr.reportFailedStatus()
 }
 

pkg/replication/finalizers.go

@@ -22,80 +22,83 @@ package replication
 
 import (
 	"context"
+	"fmt"
+	"strconv"
 	"time"
 
 	meta "k8s.io/apimachinery/pkg/apis/meta/v1"
 
 	"github.com/arangodb/arangosync-client/client"
+	"github.com/arangodb/go-driver"
 
 	api "github.com/arangodb/kube-arangodb/pkg/apis/replication/v1"
 	"github.com/arangodb/kube-arangodb/pkg/generated/clientset/versioned"
+	"github.com/arangodb/kube-arangodb/pkg/util"
 	"github.com/arangodb/kube-arangodb/pkg/util/constants"
 	"github.com/arangodb/kube-arangodb/pkg/util/errors"
 	"github.com/arangodb/kube-arangodb/pkg/util/k8sutil"
 )
 
 const (
-	maxCancelFailures = 5 // After this amount of failed cancel-synchronization attempts, the operator switch to abort-sychronization.
+	CancellationTimeout = time.Minute * 15
+	AbortTimeout        = time.Minute * 2
 )
 
-// addFinalizers adds a stop-sync finalizer to the api object when needed.
-func (dr *DeploymentReplication) addFinalizers() error {
-	apiObject := dr.apiObject
-	if apiObject.GetDeletionTimestamp() != nil {
+// addFinalizer adds new finalizer if it does not exist.
+func (dr *DeploymentReplication) addFinalizer(finalizer string) error {
+	if dr.apiObject.GetDeletionTimestamp() != nil {
 		// Delete already triggered, cannot add.
 		return nil
 	}
-	for _, f := range apiObject.GetFinalizers() {
-		if f == constants.FinalizerDeplReplStopSync {
-			// Finalizer already added
-			return nil
+	apiObject := dr.apiObject
+
+	if !finalizerExists(apiObject, finalizer) {
+		apiObject.SetFinalizers(append(apiObject.GetFinalizers(), finalizer))
+		if err := dr.updateCRSpec(apiObject.Spec); err != nil {
+			return errors.WithMessage(err, "Failed to update CR Spec")
 		}
 	}
-	apiObject.SetFinalizers(append(apiObject.GetFinalizers(), constants.FinalizerDeplReplStopSync))
-	if err := dr.updateCRSpec(apiObject.Spec); err != nil {
-		return errors.WithStack(err)
-	}
+
 	return nil
 }
 
-// runFinalizers goes through the list of ArangoDeploymentReplication finalizers to see if they can be removed.
-func (dr *DeploymentReplication) runFinalizers(ctx context.Context, p *api.ArangoDeploymentReplication) error {
-	log := dr.log.Str("replication-name", p.GetName())
-	var removalList []string
-	for _, f := range p.ObjectMeta.GetFinalizers() {
-		switch f {
-		case constants.FinalizerDeplReplStopSync:
-			log.Debug("Inspecting stop-sync finalizer")
-			if err := dr.inspectFinalizerDeplReplStopSync(ctx, p); err == nil {
-				removalList = append(removalList, f)
-			} else {
-				log.Err(err).Str("finalizer", f).Debug("Cannot remove finalizer yet")
-			}
-		}
-	}
-	// Remove finalizers (if needed)
-	if len(removalList) > 0 {
-		ignoreNotFound := false
-		if err := removeDeploymentReplicationFinalizers(dr.deps.Client.Arango(), p, removalList, ignoreNotFound); err != nil {
-			log.Err(err).Debug("Failed to update deployment replication (to remove finalizers)")
-			return errors.WithStack(err)
-		}
-	}
-	return nil
+// addFinalizers adds a required finalizers to the api object when needed.
+func (dr *DeploymentReplication) addFinalizers() error {
+	// Add stop sync replication finalizer automatically.
+	return dr.addFinalizer(constants.FinalizerDeplReplStopSync)
 }
 
-// inspectFinalizerDeplReplStopSync checks the finalizer condition for stop-sync.
-// It returns nil if the finalizer can be removed.
-func (dr *DeploymentReplication) inspectFinalizerDeplReplStopSync(ctx context.Context, p *api.ArangoDeploymentReplication) error {
-	// Inspect phase
-	if p.Status.Phase.IsFailed() {
-		dr.log.Debug("Deployment replication is already failed, safe to remove stop-sync finalizer")
-		return nil
+// runFinalizers removes stop sync finalizer if it is possible.
+func (dr *DeploymentReplication) runFinalizers(ctx context.Context, p *api.ArangoDeploymentReplication) (bool, error) {
+	if !finalizerExists(p, constants.FinalizerDeplReplStopSync) {
+		return false, nil
 	}
 
-	// Inspect deployment deletion state in source
-	abort := dr.status.CancelFailures > maxCancelFailures
+	dr.log.Str("replication-name", p.GetName()).Debug("Inspecting stop-sync finalizer")
+	if retrySoon, err := dr.inspectFinalizerDeplReplStopSync(ctx, p); err != nil {
+		return true, errors.WithMessagef(err, "Cannot remove finalizer \"%s\" yet", constants.FinalizerDeplReplStopSync)
+	} else if retrySoon {
+		// No error, but not finished. Try to reconcile soon.
+		dr.log.Debug("Synchronization is still cancelling")
+		return true, nil
+	}
+
+	removalList := []string{constants.FinalizerDeplReplStopSync}
+	if err := removeDeploymentReplicationFinalizers(dr.deps.Client.Arango(), p, removalList, false); err != nil {
+		return true, errors.WithMessage(err, "Failed to update deployment replication (to remove finalizers)")
+	}
+
+	return false, nil
+}
+
+// inspectFinalizerDeplReplStopSync checks cancellation progress.
+// When true is returned then function can be called after a few seconds to check progress.
+// When it returns false and nil error then cancellation process is done.
+func (dr *DeploymentReplication) inspectFinalizerDeplReplStopSync(ctx context.Context,
+	p *api.ArangoDeploymentReplication) (bool, error) {
+
+	abort := isTimeExceeded(p.GetDeletionTimestamp(), CancellationTimeout)
+	// Inspect deployment deletion state in source.
 	depls := dr.deps.Client.Arango().DatabaseV1().ArangoDeployments(p.GetNamespace())
 	if name := p.Spec.Source.GetDeploymentName(); name != "" {
 		depl, err := depls.Get(context.Background(), name, meta.GetOptions{})
@@ -104,7 +107,7 @@ func (dr *DeploymentReplication) inspectFinalizerDeplReplStopSync(ctx context.Co
 			abort = true
 		} else if err != nil {
 			dr.log.Err(err).Warn("Failed to get source deployment")
-			return errors.WithStack(err)
+			return false, errors.WithStack(err)
 		} else if depl.GetDeletionTimestamp() != nil {
 			dr.log.Debug("Source deployment is being deleted. Abort enabled")
 			abort = true
@@ -119,8 +122,8 @@ func (dr *DeploymentReplication) inspectFinalizerDeplReplStopSync(ctx context.Co
 			dr.log.Debug("Destination deployment is gone. Source cleanup enabled")
 			cleanupSource = true
 		} else if err != nil {
-			dr.log.Err(err).Warn("Failed to get destinaton deployment")
-			return errors.WithStack(err)
+			dr.log.Err(err).Warn("Failed to get destination deployment")
+			return false, errors.WithStack(err)
 		} else if depl.GetDeletionTimestamp() != nil {
 			dr.log.Debug("Destination deployment is being deleted. Source cleanup enabled")
 			cleanupSource = true
@@ -136,31 +139,96 @@ func (dr *DeploymentReplication) inspectFinalizerDeplReplStopSync(ctx context.Co
 			return errors.WithStack(err)
 		}*/
 		//sourceClient.Master().C
-		return errors.WithStack(errors.Newf("TODO"))
-	} else {
-		// Destination still exists, stop/abort sync
-		destClient, err := dr.createSyncMasterClient(p.Spec.Destination)
-		if err != nil {
-			dr.log.Err(err).Warn("Failed to create destination client")
-			return errors.WithStack(err)
-		}
-		req := client.CancelSynchronizationRequest{
-			WaitTimeout:  time.Minute * 3,
-			Force:        abort,
-			ForceTimeout: time.Minute * 2,
-		}
-		dr.log.Bool("abort", abort).Debug("Stopping synchronization...")
-		_, err = destClient.Master().CancelSynchronization(ctx, req)
-		if err != nil && !client.IsPreconditionFailed(err) {
-			dr.log.Err(err).Bool("abort", abort).Warn("Failed to stop synchronization")
-			dr.status.CancelFailures++
-			if err := dr.updateCRStatus(); err != nil {
-				dr.log.Err(err).Warn("Failed to update status to reflect cancel-failures increment")
-			}
-			return errors.WithStack(err)
-		}
-		return nil
+		return false, errors.WithStack(errors.Newf("TODO"))
 	}
+
+	// Destination still exists, stop/abort sync.
+	// Create a client to the destination sync master.
+	destClient, err := dr.createSyncMasterClient(p.Spec.Destination)
+	if err != nil {
+		return false, errors.WithMessage(err, "Failed to create destination synchronization master client")
+	}
+
+	// Get status from sync master.
+	syncInfo, err := destClient.Master().Status(ctx)
+	if err != nil {
+		return false, errors.WithMessage(err, "Failed to get status from target master")
+	}
+
+	// Check progress of a cancellation.
+	if syncStatus, err := dr.getCancellationProgress(syncInfo); err != nil {
+		return false, err
+	} else if syncStatus == client.SyncStatusInactive {
+		return false, nil
+	} else if syncStatus == client.SyncStatusFailed {
+		return false, errors.WithMessagef(err, "unexpected synchronization status \"%s\"", syncStatus)
+	} else if syncStatus == client.SyncStatusCancelling {
+		// Synchronization is cancelling, so request was already sent.
+		if !abort {
+			return true, nil
+		}
+
+		changed := dr.status.Conditions.Update(api.ConditionTypeAborted, abort, "Cancellation type",
+			"Cancellation will wait for source data center to be canceled with a timeout")
+		if !changed {
+			return true, nil
+		}
+		// A Request must be sent once again because abort option has changed.
+	}
+
+	// Check whether data consistency must be ensured.
+	if syncInfo.Status.IsActive() && util.BoolOrDefault(p.Spec.Cancellation.EnsureInSync, true) {
+		if inSync, inSyncShards, totalShards, err := dr.ensureInSync(ctx, destClient); err != nil {
+			return false, err
+		} else if !inSync {
+			if time.Since(dr.lastLog) > time.Second*5 {
+				dr.lastLog = time.Now()
+				dr.log.Info("Consistency is being checked, %d of %d shards are in-sync", inSyncShards, totalShards)
+			}
+
+			// Retry soon.
+			return true, nil
+		}
+	}
+
+	// From here on this code should be launched only once unless abort option is changed
+	// or replication is not in cancelling state.
+	sourceServerMode := driver.ServerModeDefault
+	if util.BoolOrDefault(p.Spec.Cancellation.SourceReadOnly) {
+		sourceServerMode = driver.ServerModeReadOnly
+	}
+	req := client.CancelSynchronizationRequest{
+		Force:            abort,
+		ForceTimeout:     AbortTimeout,
+		SourceServerMode: sourceServerMode,
+	}
+	dr.log.Interface("request", req).Info("Stopping synchronization...")
+	_, errCancel := destClient.Master().CancelSynchronization(ctx, req)
+	if errCancel != nil {
+		dr.status.Reason = fmt.Sprintf("Failed to stop synchronization: %s. Abort: %s", err.Error(), strconv.FormatBool(abort))
+	} else {
+		dr.status.Reason = "Stopping synchronization started"
+	}
+
+	// Update CR status.
+	if err := dr.updateCRStatus(); err != nil {
+		dr.log.Err(err).Warn("Failed to update replication status")
+		// Don't return with this error because original error must be returned.
+		// Not a big deal, because only reason was not saved.
+		// It will be saved on next updateCRStatus call, because reason is kept in status memory
+	}
+
+	// If err is nil then nil will be returned.
+	if errCancel != nil {
+		if abort {
+			return false, errors.WithMessage(errCancel, "Failed to abort synchronization")
+		}
+
+		return false, errors.WithMessage(errCancel, "Failed to stop synchronization")
+	}
+
+	return true, nil
+
 }
 
 // removeDeploymentReplicationFinalizers removes the given finalizers from the given DeploymentReplication.
@@ -187,3 +255,81 @@ func removeDeploymentReplicationFinalizers(crcli versioned.Interface, p *api.Ara
 	}
 	return nil
 }
+
+// finalizerExists returns true if a given finalizer exists.
+func finalizerExists(p *api.ArangoDeploymentReplication, finalizer string) bool {
+	for _, f := range p.ObjectMeta.GetFinalizers() {
+		if f == finalizer {
+			return true
+		}
+	}
+
+	return false
+}
+
+func (dr *DeploymentReplication) getCancellationProgress(syncInfo client.SyncInfo) (client.SyncStatus, error) {
+	if syncInfo.IsInactive() {
+		if len(syncInfo.Source) > 0 {
+			return "", errors.New("Inactive target data center is still configured with the endpoint set to a source DC")
+		}
+		return client.SyncStatusInactive, nil
+	}
+
+	if syncInfo.Status == client.SyncStatusInactive {
+		// There are some not finished shards but status is inactive, so it was na canceled.
+		return "", errors.New("Target data center is inactive but some shards are not closed")
+	}
+
+	return syncInfo.Status, nil
+}
+
+// ensureInSync checks whether data is consistent on both data centers.
+// During this check both data centers will be in read-only mode.
+// Return nil when data is consistent or when consistency was already checked.
+func (dr *DeploymentReplication) ensureInSync(ctx context.Context, c client.API) (bool, int, int, error) {
+	if dr.status.Conditions.IsTrue(api.ConditionTypeEnsuredInSync) {
+		return true, 0, 0, nil
+	}
+
+	cancelStatus, err := c.Master().GetSynchronizationBarrierStatus(ctx)
+	if err != nil {
+		return false, 0, 0, errors.WithMessage(err, "Can not get synchronization barrier status")
+	}
+
+	if !cancelStatus.SourceServerReadonly ||
+		dr.status.Conditions.Update(api.ConditionTypeEnsuredInSync, false, "Consistent", "Data on both data centers is not the same") {
+		// If `GetSynchronizationBarrierStatus` could return active barrier then it would not create the above condition.
+		if err := c.Master().CreateSynchronizationBarrier(ctx); err != nil {
+			if driver.IsPreconditionFailed(err) {
+				dr.log.Info("Can not create synchronization barrier because synchronization is not running")
+				return false, 0, 0, nil
+			}
+
+			return false, 0, 0, errors.WithMessage(err, "Can not create synchronization barrier")
+		}
+
+		if err := dr.updateCRStatus(); err != nil {
+			return false, 0, 0, errors.WithMessage(err, "Failed to update ArangoDeploymentReplication status")
+		}
+
+		dr.log.Info("Synchronization barrier created, both data centers are in read-only mode")
+	}
+
+	totalShards := cancelStatus.InSyncShards + cancelStatus.NotInSyncShards
+	if cancelStatus.InSyncShards > 0 && cancelStatus.NotInSyncShards == 0 {
+		if dr.status.Conditions.Update(api.ConditionTypeEnsuredInSync, true, "Consistent", "Data on both data centers is the same") {
+			if err := dr.updateCRStatus(); err != nil {
+				return false, 0, 0, errors.WithMessage(err, "Failed to update ArangoDeploymentReplication status")
+			}
+		}
+
+		return true, cancelStatus.InSyncShards, totalShards, nil
+	}
+
+	return false, cancelStatus.InSyncShards, totalShards, nil
+}
+
+// isTimeExceeded returns true when a time exceeds a given timeout.
+func isTimeExceeded(t *meta.Time, timeout time.Duration) bool {
+	return t != nil && time.Since(t.Time) > timeout
+}

pkg/replication/sync_inspector.go

@@ -54,18 +54,27 @@ func (dr *DeploymentReplication) inspectDeploymentReplication(lastInterval time.
 	}
 
 	// Is the deployment in failed state, if so, give up.
-	if dr.status.Phase == api.DeploymentReplicationPhaseFailed {
+	if dr.status.Phase.IsFailed() {
 		dr.log.Debug("Deployment replication is in Failed state.")
 		return nextInterval
 	}
 
 	// Is delete triggered?
-	if dr.apiObject.GetDeletionTimestamp() != nil {
-		// Deployment replication is triggered for deletion.
-		if err := dr.runFinalizers(ctx, dr.apiObject); err != nil {
-			dr.log.Err(err).Warn("Failed to run finalizers")
-			hasError = true
+	if timestamp := dr.apiObject.GetDeletionTimestamp(); timestamp != nil {
+		// Resource is being deleted.
+		retrySoon, err := dr.runFinalizers(ctx, dr.apiObject)
+		if err != nil || retrySoon {
+			if err != nil {
+				dr.log.Err(err).Warn("Failed to run finalizers")
+			}
+			timeout := CancellationTimeout + AbortTimeout
+			if isTimeExceeded(timestamp, timeout) {
+				// Cancellation and abort timeout exceeded, so it must go into failed state.
+				dr.failOnError(err, fmt.Sprintf("Failed to cancel synchronization in %s", timeout.String()))
+			}
 		}
+
+		return cancellationInterval
 	} else {
 		// Inspect configuration status
 		destClient, err := dr.createSyncMasterClient(spec.Destination)
@@ -98,15 +107,19 @@ func (dr *DeploymentReplication) inspectDeploymentReplication(lastInterval time.
 					} else {
 						if isIncomingEndpoint {
 							// Destination is correctly configured
-							dr.status.Conditions.Update(api.ConditionTypeConfigured, true, "Active", "Destination syncmaster is configured correctly and active")
+
+							dr.status.Conditions.Update(api.ConditionTypeConfigured, true, api.ConditionConfiguredReasonActive,
+								"Destination syncmaster is configured correctly and active")
 							dr.status.Destination = createEndpointStatus(destStatus, "")
-							dr.status.IncomingSynchronization = dr.inspectIncomingSynchronizationStatus(ctx, destClient, driver.Version(destArangosyncVersion.Version), destStatus.Shards)
+							dr.status.IncomingSynchronization = dr.inspectIncomingSynchronizationStatus(ctx, destClient,
+								driver.Version(destArangosyncVersion.Version), destStatus.Shards)
 							updateStatusNeeded = true
 						} else {
 							// Sync is active, but from different source
 							dr.log.Warn("Destination syncmaster is configured for different source")
 							cancelSyncNeeded = true
-							if dr.status.Conditions.Update(api.ConditionTypeConfigured, false, "Invalid", "Destination syncmaster is configured for different source") {
+							if dr.status.Conditions.Update(api.ConditionTypeConfigured, false, api.ConditionConfiguredReasonInvalid,
+								"Destination syncmaster is configured for different source") {
 								updateStatusNeeded = true
 							}
 						}
@@ -114,7 +127,8 @@ func (dr *DeploymentReplication) inspectDeploymentReplication(lastInterval time.
 				} else {
 					// Destination has correct source, but is inactive
 					configureSyncNeeded = true
-					if dr.status.Conditions.Update(api.ConditionTypeConfigured, false, "Inactive", "Destination syncmaster is configured correctly but in-active") {
+					if dr.status.Conditions.Update(api.ConditionTypeConfigured, false, api.ConditionConfiguredReasonInactive,
+						"Destination syncmaster is configured correctly but in-active") {
 						updateStatusNeeded = true
 					}
 				}