DRA kubelet: refactor gRPC call timeouts

Some of the E2E node tests were flaky. Their timeout apparently was chosen
under the assumption that kubelet would retry immediately after a failed gRPC
call, with a factor of 2 as safety margin. But according to
kubernetes@0449cef,
kubelet has a different, higher retry period of 90 seconds, which was exactly
the test timeout. The test timeout has to be higher than that.

As the tests don't use the gRPC call timeout anymore, it can be made
private. While at it, the name and documentation gets updated.

Author: pohly

pkg/kubelet/cm/dra/plugin/client.go

@@ -33,8 +33,6 @@ import (
 	drapb "k8s.io/kubelet/pkg/apis/dra/v1alpha4"
 )
 
-const PluginClientTimeout = 45 * time.Second
-
 // NewDRAPluginClient returns a wrapper around those gRPC methods of a DRA
 // driver kubelet plugin which need to be called by kubelet. The wrapper
 // handles gRPC connection management and logging. Connections are reused
@@ -60,7 +58,7 @@ type Plugin struct {
 	conn                    *grpc.ClientConn
 	endpoint                string
 	highestSupportedVersion *utilversion.Version
-	clientTimeout           time.Duration
+	clientCallTimeout       time.Duration
 }
 
 func (p *Plugin) getOrCreateGRPCConn() (*grpc.ClientConn, error) {
@@ -116,7 +114,7 @@ func (p *Plugin) NodePrepareResources(
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(ctx, p.clientTimeout)
+	ctx, cancel := context.WithTimeout(ctx, p.clientCallTimeout)
 	defer cancel()
 
 	nodeClient := drapb.NewNodeClient(conn)
@@ -138,7 +136,7 @@ func (p *Plugin) NodeUnprepareResources(
 		return nil, err
 	}
 
-	ctx, cancel := context.WithTimeout(ctx, p.clientTimeout)
+	ctx, cancel := context.WithTimeout(ctx, p.clientCallTimeout)
 	defer cancel()
 
 	nodeClient := drapb.NewNodeClient(conn)

pkg/kubelet/cm/dra/plugin/client_test.go

@@ -243,9 +243,9 @@ func TestNodeUnprepareResources(t *testing.T) {
 			defer teardown()
 
 			p := &Plugin{
-				backgroundCtx: ctx,
-				endpoint:      addr,
-				clientTimeout: PluginClientTimeout,
+				backgroundCtx:     ctx,
+				endpoint:          addr,
+				clientCallTimeout: defaultClientCallTimeout,
 			}
 
 			conn, err := p.getOrCreateGRPCConn()

pkg/kubelet/cm/dra/plugin/plugin.go

@@ -35,6 +35,15 @@ import (
 	"k8s.io/kubernetes/pkg/kubelet/pluginmanager/cache"
 )
 
+// defaultClientCallTimeout is the default amount of time that a DRA driver has
+// to respond to any of the gRPC calls. kubelet uses this value by passing nil
+// to RegisterPlugin. Some tests use a different, usually shorter timeout to
+// speed up testing.
+//
+// This is half of the kubelet retry period (according to
+// https://github.com/kubernetes/kubernetes/commit/0449cef8fd5217d394c5cd331d852bd50983e6b3).
+const defaultClientCallTimeout = 45 * time.Second
+
 // RegistrationHandler is the handler which is fed to the pluginwatcher API.
 type RegistrationHandler struct {
 	// backgroundCtx is used for all future activities of the handler.
@@ -144,7 +153,7 @@ func (h *RegistrationHandler) RegisterPlugin(pluginName string, endpoint string,
 
 	var timeout time.Duration
 	if pluginClientTimeout == nil {
-		timeout = PluginClientTimeout
+		timeout = defaultClientCallTimeout
 	} else {
 		timeout = *pluginClientTimeout
 	}
@@ -157,7 +166,7 @@ func (h *RegistrationHandler) RegisterPlugin(pluginName string, endpoint string,
 		conn:                    nil,
 		endpoint:                endpoint,
 		highestSupportedVersion: highestSupportedVersion,
-		clientTimeout:           timeout,
+		clientCallTimeout:       timeout,
 	}
 
 	// Storing endpoint of newly registered DRA Plugin into the map, where plugin name will be the key

staging/src/k8s.io/dynamic-resource-allocation/structured/allocator_test.go

@@ -137,26 +137,6 @@ func classWithConfig(name, driver, attribute string) *resourceapi.DeviceClass {
 	return class
 }
 
-// generate a DeviceClass object with the given name and the node selector
-// that selects nodes with the region label set to either "west" or "east".
-func classWithSuitableNodes(name, driver string) *resourceapi.DeviceClass {
-	class := class(name, driver)
-	class.Spec.SuitableNodes = &v1.NodeSelector{
-		NodeSelectorTerms: []v1.NodeSelectorTerm{
-			{
-				MatchExpressions: []v1.NodeSelectorRequirement{
-					{
-						Key:      regionKey,
-						Operator: v1.NodeSelectorOpIn,
-						Values:   []string{region1, region2},
-					},
-				},
-			},
-		},
-	}
-	return class
-}
-
 // generate a ResourceClaim object with the given name and device requests.
 func claimWithRequests(name string, constraints []resourceapi.DeviceConstraint, requests ...resourceapi.DeviceRequest) *resourceapi.ResourceClaim {
 	return &resourceapi.ResourceClaim{