Merge pull request #121065 from byako/issue121063

DRA: notify ResourceClaim or ResourceClass of params error

staging/src/k8s.io/dynamic-resource-allocation/controller/controller.go

@@ -94,7 +94,7 @@ type Driver interface {
 	// If selectedNode is set, the driver must attempt to allocate for that
 	// node. If that is not possible, it must return an error. The
 	// controller will call UnsuitableNodes and pass the new information to
-	// the scheduler, which then will lead to selecting a diffent node
+	// the scheduler, which then will lead to selecting a different node
 	// if the current one is not suitable.
 	//
 	// The Claim, ClaimParameters, Class, ClassParameters fields of "claims" parameter
@@ -357,7 +357,6 @@ func (ctrl *controller) Run(workers int) {
 var errRequeue = errors.New("requeue")
 
 // errPeriodic is a special error instance that functions can return
-// to request silent instance that functions can return
 // to request silent retrying at a fixed rate.
 var errPeriodic = errors.New("periodic")
 
@@ -536,8 +535,9 @@ func (ctrl *controller) syncClaim(ctx context.Context, claim *resourcev1alpha2.R
 		return errRequeue
 	}
 
-	// Check parameters.
+	// Check parameters. Do not record event to Claim if its parameters are invalid,
-	claimParameters, classParameters, err := ctrl.getParameters(ctx, claim, class)
+	// syncKey will record the error.
+	claimParameters, classParameters, err := ctrl.getParameters(ctx, claim, class, false)
 	if err != nil {
 		return err
 	}
@@ -558,14 +558,18 @@ func (ctrl *controller) syncClaim(ctx context.Context, claim *resourcev1alpha2.R
 	return nil
 }
 
-func (ctrl *controller) getParameters(ctx context.Context, claim *resourcev1alpha2.ResourceClaim, class *resourcev1alpha2.ResourceClass) (claimParameters, classParameters interface{}, err error) {
+func (ctrl *controller) getParameters(ctx context.Context, claim *resourcev1alpha2.ResourceClaim, class *resourcev1alpha2.ResourceClass, notifyClaim bool) (claimParameters, classParameters interface{}, err error) {
 	classParameters, err = ctrl.driver.GetClassParameters(ctx, class)
 	if err != nil {
+		ctrl.eventRecorder.Event(class, v1.EventTypeWarning, "Failed", err.Error())
 		err = fmt.Errorf("class parameters %s: %v", class.ParametersRef, err)
 		return
 	}
 	claimParameters, err = ctrl.driver.GetClaimParameters(ctx, claim, class, classParameters)
 	if err != nil {
+		if notifyClaim {
+			ctrl.eventRecorder.Event(claim, v1.EventTypeWarning, "Failed", err.Error())
+		}
 		err = fmt.Errorf("claim parameters %s: %v", claim.Spec.ParametersRef, err)
 		return
 	}
@@ -688,9 +692,10 @@ func (ctrl *controller) checkPodClaim(ctx context.Context, pod *v1.Pod, podClaim
 	if class.DriverName != ctrl.name {
 		return nil, nil
 	}
-	// Check parameters.
+	// Check parameters. Record event to claim and pod if parameters are invalid.
-	claimParameters, classParameters, err := ctrl.getParameters(ctx, claim, class)
+	claimParameters, classParameters, err := ctrl.getParameters(ctx, claim, class, true)
 	if err != nil {
+		ctrl.eventRecorder.Event(pod, v1.EventTypeWarning, "Failed", fmt.Sprintf("claim %v: %v", claim.Name, err.Error()))
 		return nil, err
 	}
 	return &ClaimAllocation{

staging/src/k8s.io/kubelet/pkg/apis/dra/v1alpha3/api.pb.go

@@ -482,7 +482,7 @@ const _ = grpc.SupportPackageIsVersion4
 type NodeClient interface {
 	// NodePrepareResources prepares several ResourceClaims
 	// for use on the node. If an error is returned, the
-	// response is ignored. Failures for individidual claims
+	// response is ignored. Failures for individual claims
 	// can be reported inside NodePrepareResourcesResponse.
 	NodePrepareResources(ctx context.Context, in *NodePrepareResourcesRequest, opts ...grpc.CallOption) (*NodePrepareResourcesResponse, error)
 	// NodeUnprepareResources is the opposite of NodePrepareResources.
@@ -520,7 +520,7 @@ func (c *nodeClient) NodeUnprepareResources(ctx context.Context, in *NodeUnprepa
 type NodeServer interface {
 	// NodePrepareResources prepares several ResourceClaims
 	// for use on the node. If an error is returned, the
-	// response is ignored. Failures for individidual claims
+	// response is ignored. Failures for individual claims
 	// can be reported inside NodePrepareResourcesResponse.
 	NodePrepareResources(context.Context, *NodePrepareResourcesRequest) (*NodePrepareResourcesResponse, error)
 	// NodeUnprepareResources is the opposite of NodePrepareResources.

staging/src/k8s.io/kubelet/pkg/apis/dra/v1alpha3/api.proto

@@ -34,7 +34,7 @@ option (gogoproto.goproto_unrecognized_all) = false;
 service Node {
   // NodePrepareResources prepares several ResourceClaims
   // for use on the node. If an error is returned, the
-  // response is ignored. Failures for individidual claims
+  // response is ignored. Failures for individual claims
   // can be reported inside NodePrepareResourcesResponse.
   rpc NodePrepareResources (NodePrepareResourcesRequest)
     returns (NodePrepareResourcesResponse) {}