Merge pull request #22392 from janetkuo/retry-update-precondition

pkg/util/deployment/deployment.go

@@ -29,6 +29,7 @@ import (
 	clientset "k8s.io/kubernetes/pkg/client/clientset_generated/internalclientset"
 	"k8s.io/kubernetes/pkg/controller"
 	"k8s.io/kubernetes/pkg/labels"
+	"k8s.io/kubernetes/pkg/util/errors"
 	"k8s.io/kubernetes/pkg/util/integer"
 	intstrutil "k8s.io/kubernetes/pkg/util/intstr"
 	labelsutil "k8s.io/kubernetes/pkg/util/labels"
@@ -188,27 +189,32 @@ func addHashKeyToRSAndPods(deployment *extensions.Deployment, c clientset.Interf
 	}))
 	rsUpdated := false
 	// 1. Add hash template label to the rs. This ensures that any newly created pods will have the new label.
-	if len(updatedRS.Spec.Template.Labels[extensions.DefaultDeploymentUniqueLabelKey]) == 0 {
-		updatedRS, rsUpdated, err = rsutil.UpdateRSWithRetries(c.Extensions().ReplicaSets(namespace), updatedRS, func(updated *extensions.ReplicaSet) {
+	updatedRS, rsUpdated, err = rsutil.UpdateRSWithRetries(c.Extensions().ReplicaSets(namespace), updatedRS,
+		func(updated *extensions.ReplicaSet) error {
+			// Precondition: the RS doesn't contain the new hash in its pod template label.
+			if updated.Spec.Template.Labels[extensions.DefaultDeploymentUniqueLabelKey] == hash {
+				return errors.ErrPreconditionViolated
+			}
 			updated.Spec.Template.Labels = labelsutil.AddLabel(updated.Spec.Template.Labels, extensions.DefaultDeploymentUniqueLabelKey, hash)
+			return nil
 		})
-		if err != nil {
-			return nil, fmt.Errorf("error updating %s %s/%s pod template label with template hash: %v", updatedRS.Kind, updatedRS.Namespace, updatedRS.Name, err)
-		}
-		if rsUpdated {
-			// Make sure rs pod template is updated so that it won't create pods without the new label (orphaned pods).
-			if updatedRS.Generation > updatedRS.Status.ObservedGeneration {
-				if err = waitForReplicaSetUpdated(c, updatedRS.Generation, namespace, updatedRS.Name); err != nil {
-					return nil, fmt.Errorf("error waiting for %s %s/%s generation %d observed by controller: %v", updatedRS.Kind, updatedRS.Namespace, updatedRS.Name, updatedRS.Generation, err)
-				}
+	if err != nil {
+		return nil, fmt.Errorf("error updating %s %s/%s pod template label with template hash: %v", updatedRS.Kind, updatedRS.Namespace, updatedRS.Name, err)
+	}
+	if rsUpdated {
+		// Make sure rs pod template is updated so that it won't create pods without the new label (orphaned pods).
+		if updatedRS.Generation > updatedRS.Status.ObservedGeneration {
+			if err = waitForReplicaSetUpdated(c, updatedRS.Generation, namespace, updatedRS.Name); err != nil {
+				return nil, fmt.Errorf("error waiting for %s %s/%s generation %d observed by controller: %v", updatedRS.Kind, updatedRS.Namespace, updatedRS.Name, updatedRS.Generation, err)
 			}
-			glog.V(4).Infof("Observed the update of %s %s/%s's pod template with hash %s.", rs.Kind, rs.Namespace, rs.Name, hash)
-		} else {
-			// If RS wasn't updated but didn't return error in step 1, we've hit a RS not found error.
-			// Return here and retry in the next sync loop.
-			return &rs, nil
 		}
+		glog.V(4).Infof("Observed the update of %s %s/%s's pod template with hash %s.", rs.Kind, rs.Namespace, rs.Name, hash)
+	} else {
+		// If RS wasn't updated but didn't return error in step 1, we've hit a RS not found error.
+		// Return here and retry in the next sync loop.
+		return &rs, nil
 	}
+	glog.V(4).Infof("Observed the update of rs %s's pod template with hash %s.", rs.Name, hash)
 
 	// 2. Update all pods managed by the rs to have the new hash label, so they will be correctly adopted.
 	selector, err := unversioned.LabelSelectorAsSelector(updatedRS.Spec.Selector)
@@ -232,10 +238,16 @@ func addHashKeyToRSAndPods(deployment *extensions.Deployment, c clientset.Interf
 
 	// 3. Update rs label and selector to include the new hash label
 	// Copy the old selector, so that we can scrub out any orphaned pods
-	if updatedRS, rsUpdated, err = rsutil.UpdateRSWithRetries(c.Extensions().ReplicaSets(namespace), updatedRS, func(updated *extensions.ReplicaSet) {
-		updated.Labels = labelsutil.AddLabel(updated.Labels, extensions.DefaultDeploymentUniqueLabelKey, hash)
-		updated.Spec.Selector = labelsutil.AddLabelToSelector(updated.Spec.Selector, extensions.DefaultDeploymentUniqueLabelKey, hash)
-	}); err != nil {
+	if updatedRS, rsUpdated, err = rsutil.UpdateRSWithRetries(c.Extensions().ReplicaSets(namespace), updatedRS,
+		func(updated *extensions.ReplicaSet) error {
+			// Precondition: the RS doesn't contain the new hash in its label or selector.
+			if updated.Labels[extensions.DefaultDeploymentUniqueLabelKey] == hash && updated.Spec.Selector.MatchLabels[extensions.DefaultDeploymentUniqueLabelKey] == hash {
+				return errors.ErrPreconditionViolated
+			}
+			updated.Labels = labelsutil.AddLabel(updated.Labels, extensions.DefaultDeploymentUniqueLabelKey, hash)
+			updated.Spec.Selector = labelsutil.AddLabelToSelector(updated.Spec.Selector, extensions.DefaultDeploymentUniqueLabelKey, hash)
+			return nil
+		}); err != nil {
 		return nil, fmt.Errorf("error updating %s %s/%s label and selector with template hash: %v", updatedRS.Kind, updatedRS.Namespace, updatedRS.Name, err)
 	}
 	if rsUpdated {
@@ -265,14 +277,20 @@ func labelPodsWithHash(podList *api.PodList, rs *extensions.ReplicaSet, c client
 	for _, pod := range podList.Items {
 		// Only label the pod that doesn't already have the new hash
 		if pod.Labels[extensions.DefaultDeploymentUniqueLabelKey] != hash {
-			if _, podUpdated, err := podutil.UpdatePodWithRetries(c.Core().Pods(namespace), &pod, func(podToUpdate *api.Pod) {
-				podToUpdate.Labels = labelsutil.AddLabel(podToUpdate.Labels, extensions.DefaultDeploymentUniqueLabelKey, hash)
-			}); err != nil {
+			if _, podUpdated, err := podutil.UpdatePodWithRetries(c.Core().Pods(namespace), &pod,
+				func(podToUpdate *api.Pod) error {
+					// Precondition: the pod doesn't contain the new hash in its label.
+					if podToUpdate.Labels[extensions.DefaultDeploymentUniqueLabelKey] == hash {
+						return errors.ErrPreconditionViolated
+					}
+					podToUpdate.Labels = labelsutil.AddLabel(podToUpdate.Labels, extensions.DefaultDeploymentUniqueLabelKey, hash)
+					return nil
+				}); err != nil {
 				return false, fmt.Errorf("error in adding template hash label %s to pod %+v: %s", hash, pod, err)
 			} else if podUpdated {
 				glog.V(4).Infof("Labeled %s %s/%s of %s %s/%s with hash %s.", pod.Kind, pod.Namespace, pod.Name, rs.Kind, rs.Namespace, rs.Name, hash)
 			} else {
-				// If the pod wasn't updated but didn't return error when we try to update it, we've hit a pod not found error.
+				// If the pod wasn't updated but didn't return error when we try to update it, we've hit "pod not found" or "precondition violated" error.
 				// Then we can't say all pods are labeled
 				allPodsLabeled = false
 			}

pkg/util/errors/errors.go

@@ -16,7 +16,10 @@ limitations under the License.
 
 package errors
 
-import "fmt"
+import (
+	"errors"
+	"fmt"
+)
 
 // Aggregate represents an object that contains multiple errors, but does not
 // necessarily have singular semantic meaning.
@@ -148,3 +151,6 @@ func AggregateGoroutines(funcs ...func() error) Aggregate {
 	}
 	return NewAggregate(errs)
 }
+
+// ErrPreconditionViolated is returned when the precondition is violated
+var ErrPreconditionViolated = errors.New("precondition is violated")

pkg/util/pod/pod.go

@@ -26,6 +26,7 @@ import (
 	"k8s.io/kubernetes/pkg/api"
 	"k8s.io/kubernetes/pkg/api/errors"
 	unversionedcore "k8s.io/kubernetes/pkg/client/typed/generated/core/unversioned"
+	errorsutil "k8s.io/kubernetes/pkg/util/errors"
 	hashutil "k8s.io/kubernetes/pkg/util/hash"
 	"k8s.io/kubernetes/pkg/util/wait"
 )
@@ -38,7 +39,7 @@ func GetPodTemplateSpecHash(template api.PodTemplateSpec) uint32 {
 
 // TODO: use client library instead when it starts to support update retries
 //       see https://github.com/kubernetes/kubernetes/issues/21479
-type updatePodFunc func(pod *api.Pod)
+type updatePodFunc func(pod *api.Pod) error
 
 // UpdatePodWithRetries updates a pod with given applyUpdate function. Note that pod not found error is ignored.
 // The returned bool value can be used to tell if the pod is actually updated.
@@ -52,8 +53,9 @@ func UpdatePodWithRetries(podClient unversionedcore.PodInterface, pod *api.Pod,
 			return false, err
 		}
 		// Apply the update, then attempt to push it to the apiserver.
-		// TODO: add precondition for update
-		applyUpdate(pod)
+		if err = applyUpdate(pod); err != nil {
+			return false, err
+		}
 		if pod, err = podClient.Update(pod); err == nil {
 			// Update successful.
 			return true, nil
@@ -66,12 +68,22 @@ func UpdatePodWithRetries(podClient unversionedcore.PodInterface, pod *api.Pod,
 		podUpdated = true
 	}
 
+	// Handle returned error from wait poll
 	if err == wait.ErrWaitTimeout {
 		err = fmt.Errorf("timed out trying to update pod: %+v", oldPod)
 	}
+	// Ignore the pod not found error, but the pod isn't updated.
 	if errors.IsNotFound(err) {
 		glog.V(4).Infof("%s %s/%s is not found, skip updating it.", oldPod.Kind, oldPod.Namespace, oldPod.Name)
 		err = nil
 	}
+	// Ignore the precondition violated error, but the pod isn't updated.
+	if err == errorsutil.ErrPreconditionViolated {
+		glog.V(4).Infof("%s %s/%s precondition doesn't hold, skip updating it.", oldPod.Kind, oldPod.Namespace, oldPod.Name)
+		err = nil
+	}
+
+	// If the error is non-nil the returned pod cannot be trusted; if podUpdated is false, the pod isn't updated;
+	// if the error is nil and podUpdated is true, the returned pod contains the applied update.
 	return pod, podUpdated, err
 }

pkg/util/replicaset/replicaset.go

@@ -24,12 +24,13 @@ import (
 	"k8s.io/kubernetes/pkg/api/errors"
 	"k8s.io/kubernetes/pkg/apis/extensions"
 	unversionedextensions "k8s.io/kubernetes/pkg/client/typed/generated/extensions/unversioned"
+	errorsutil "k8s.io/kubernetes/pkg/util/errors"
 	"k8s.io/kubernetes/pkg/util/wait"
 )
 
 // TODO: use client library instead when it starts to support update retries
 //       see https://github.com/kubernetes/kubernetes/issues/21479
-type updateRSFunc func(rs *extensions.ReplicaSet)
+type updateRSFunc func(rs *extensions.ReplicaSet) error
 
 // UpdateRSWithRetries updates a RS with given applyUpdate function. Note that RS not found error is ignored.
 // The returned bool value can be used to tell if the RS is actually updated.
@@ -43,8 +44,9 @@ func UpdateRSWithRetries(rsClient unversionedextensions.ReplicaSetInterface, rs
 			return false, err
 		}
 		// Apply the update, then attempt to push it to the apiserver.
-		// TODO: add precondition for update
-		applyUpdate(rs)
+		if err = applyUpdate(rs); err != nil {
+			return false, err
+		}
 		if rs, err = rsClient.Update(rs); err == nil {
 			// Update successful.
 			return true, nil
@@ -57,6 +59,7 @@ func UpdateRSWithRetries(rsClient unversionedextensions.ReplicaSetInterface, rs
 		rsUpdated = true
 	}
 
+	// Handle returned error from wait poll
 	if err == wait.ErrWaitTimeout {
 		err = fmt.Errorf("timed out trying to update RS: %+v", oldRs)
 	}
@@ -65,7 +68,13 @@ func UpdateRSWithRetries(rsClient unversionedextensions.ReplicaSetInterface, rs
 		glog.V(4).Infof("%s %s/%s is not found, skip updating it.", oldRs.Kind, oldRs.Namespace, oldRs.Name)
 		err = nil
 	}
-	// If the error is non-nil the returned controller cannot be trusted, if it is nil, the returned
-	// controller contains the applied update.
+	// Ignore the precondition violated error, but the RS isn't updated.
+	if err == errorsutil.ErrPreconditionViolated {
+		glog.V(4).Infof("%s %s/%s precondition doesn't hold, skip updating it.", oldRs.Kind, oldRs.Namespace, oldRs.Name)
+		err = nil
+	}
+
+	// If the error is non-nil the returned RS cannot be trusted; if rsUpdated is false, the contoller isn't updated;
+	// if the error is nil and rsUpdated is true, the returned RS contains the applied update.
 	return rs, rsUpdated, err
 }