Check for nil waitOp before examining error code

Change-Id: Iab5a2c41528e49c5d8e2e092a22a89c2f8edd232

Author: mattcary

pkg/common/utils.go

@@ -363,7 +363,7 @@ func isGoogleAPIError(err error) (codes.Code, error) {
 }
 
 func loggedErrorForCode(msg string, code codes.Code, err error) error {
-	klog.Errorf(msg+"%v", err.Error())
+	klog.Errorf("%v: %s %v", code, msg, err.Error())
 	return status.Errorf(code, msg+"%v", err.Error())
 }
 

pkg/gce-cloud-provider/compute/gce-compute.go

@@ -670,7 +670,7 @@ func (cloud *CloudProvider) insertConstructedDisk(ctx context.Context, disk *com
 	}
 
 	if filterErr := cloud.processDiskAlreadyExistErr(ctx, err, project, volKey, params, capacityRange, multiWriter, accessMode); filterErr != nil {
-		return common.NewTemporaryError(codes.Unavailable, fmt.Errorf("unknown error when polling the operation: %w", err))
+		return common.NewTemporaryError(codes.Unavailable, fmt.Errorf("unknown error when polling the operation: %w; filter error %v", err, filterErr))
 	}
 	return nil
 }
@@ -1063,7 +1063,7 @@ func (cloud *CloudProvider) waitForZonalOp(ctx context.Context, project, opName
 	return wait.ExponentialBackoff(WaitForOpBackoff, func() (bool, error) {
 		waitOp, err := cloud.service.ZoneOperations.Wait(project, zone, opName).Context(ctx).Do()
 		// In case of service unavailable do not propogate the error so ExponentialBackoff will retry
-		if err != nil && waitOp.HttpErrorStatusCode == 503 {
+		if err != nil && waitOp != nil && waitOp.HttpErrorStatusCode == 503 {
 			klog.Errorf("WaitForZonalOp(op: %s, zone: %#v, err: %v) failed to poll the operation", opName, zone, err)
 			return false, nil
 		}
@@ -1080,7 +1080,7 @@ func (cloud *CloudProvider) waitForRegionalOp(ctx context.Context, project, opNa
 	return wait.ExponentialBackoff(WaitForOpBackoff, func() (bool, error) {
 		waitOp, err := cloud.service.RegionOperations.Wait(project, region, opName).Context(ctx).Do()
 		// In case of service unavailable do not propogate the error so ExponentialBackoff will retry
-		if err != nil && waitOp.HttpErrorStatusCode == 503 {
+		if err != nil && waitOp != nil && waitOp.HttpErrorStatusCode == 503 {
 			klog.Errorf("WaitForRegionalOp(op: %s, region: %#v, err: %v) failed to poll the operation", opName, region, err)
 			return false, nil
 		}
@@ -1089,6 +1089,10 @@ func (cloud *CloudProvider) waitForRegionalOp(ctx context.Context, project, opNa
 			return false, err
 		}
 		done, err := opIsDone(waitOp)
+		if err != nil && status.Code(err) == codes.Unavailable {
+			klog.Errorf("retriable error in op: %+v", err)
+			return false, nil
+		}
 		return done, err
 	})
 }
@@ -1097,7 +1101,7 @@ func (cloud *CloudProvider) waitForGlobalOp(ctx context.Context, project, opName
 	return wait.ExponentialBackoff(WaitForOpBackoff, func() (bool, error) {
 		waitOp, err := cloud.service.GlobalOperations.Wait(project, opName).Context(ctx).Do()
 		// In case of service unavailable do not propogate the error so ExponentialBackoff will retry
-		if err != nil && waitOp.HttpErrorStatusCode == 503 {
+		if err != nil && waitOp != nil && waitOp.HttpErrorStatusCode == 503 {
 			klog.Errorf("WaitForGlobalOp(op: %s, err: %v) failed to poll the operation", opName, err)
 			return false, nil
 		}
@@ -1197,6 +1201,7 @@ func codeForGCEOpError(err computev1.OperationErrorErrors) codes.Code {
 		"QUOTA_EXCEEDED":                            codes.ResourceExhausted,
 		"ZONE_RESOURCE_POOL_EXHAUSTED":              codes.Unavailable,
 		"ZONE_RESOURCE_POOL_EXHAUSTED_WITH_DETAILS": codes.Unavailable,
+		"INTERNAL_ERROR":                            codes.Unavailable,
 		"REGION_QUOTA_EXCEEDED":                     codes.ResourceExhausted,
 		"RATE_LIMIT_EXCEEDED":                       codes.ResourceExhausted,
 		"INVALID_USAGE":                             codes.InvalidArgument,

pkg/gce-pd-csi-driver/controller.go

@@ -632,7 +632,7 @@ func (gceCS *GCEControllerServer) createSingleDeviceDisk(ctx context.Context, re
 
 	disk, err := gceCS.createSingleDisk(ctx, req, params, volKey, zones, accessMode)
 	if err != nil {
-		return nil, common.LoggedError("CreateVolume failed: %v", err)
+		return nil, common.LoggedError("CreateVolume failed: ", err)
 	}
 
 	return gceCS.generateCreateVolumeResponseWithVolumeId(disk, zones, params, dataCacheParams, enableDataCache, volumeID), nil

test/e2e/tests/multi_zone_e2e_test.go

@@ -64,7 +64,7 @@ func checkSkipMultiZoneTests() {
 	// TODO: Remove this once hyperdisk-ml SKU is supported
 	// If you want to run these tests, set the env variable: RUN_MULTI_ZONE_TESTS=true
 	if !runMultiZoneTests() {
-		Skip("Not running multi-zone tests, as RUN_MULTI_ZONE_TESTS is falsy")
+		Skip("Not running multi-zone tests without RUN_MULTI_ZONE_TESTS=true")
 	}
 }
 
@@ -1235,8 +1235,9 @@ var _ = Describe("GCE PD CSI Driver Multi-Zone", func() {
 	})
 
 	It("Should successfully run through entire lifecycle of a HdHA volume on instances in 2 zones", func() {
-		// Create new driver and client
+		Skip("Flaking on GCP errors. Google internal bug: 463743704")
 
+		// Create new driver and client
 		Expect(hyperdiskTestContexts).NotTo(BeEmpty())
 
 		zoneToContext := map[string]*remote.TestContext{}
@@ -1255,10 +1256,12 @@ var _ = Describe("GCE PD CSI Driver Multi-Zone", func() {
 		}
 
 		Expect(len(zoneToContext)).To(Equal(2), "Must have instances in 2 zones")
+		klog.Infof("Using zones %s and %s", zones[0], zones[1])
 
 		controllerContext := zoneToContext[zones[0]]
 		controllerClient := controllerContext.Client
 		controllerInstance := controllerContext.Instance
+		klog.Infof("Using controller instance %v", controllerInstance.GetName())
 
 		p, _, _ := controllerInstance.GetIdentity()
 
@@ -1321,6 +1324,8 @@ var _ = Describe("GCE PD CSI Driver Multi-Zone", func() {
 	})
 
 	It("Should create a HdHA instance, write to it, force-attach it to another instance, and read the same data", func() {
+		Skip("Flaking on GCP errors. Google internal bug: 463743704")
+
 		Expect(hyperdiskTestContexts).NotTo(BeEmpty())
 
 		zoneToContext := map[string]*remote.TestContext{}

test/e2e/tests/setup_e2e_test.go

@@ -60,7 +60,7 @@ var (
 	// Multi-writer is only supported on M3, C3, and N4
 	// https://cloud.google.com/compute/docs/disks/sharing-disks-between-vms#hd-multi-writer
 	hdMachineType    = flag.String("hyperdisk-machine-type", "c3-standard-4", "Type of machine to provision instance on, or `none' to skip")
-	hdMinCpuPlatform = flag.String("hyperdisk-min-cpu-platform", "sapphirerapids", "Minimum CPU architecture")
+	hdMinCpuPlatform = flag.String("hyperdisk-min-cpu-platform", "Intel Sapphire Rapids", "Minimum CPU architecture")
 
 	// Some architectures don't have local ssd. Give way to opt out of tests like datacache.
 	skipLocalSsdTests = flag.Bool("skip-local-ssd-tests", false, "Skip local ssd tests like datacache")