Change behavior of repair disks to return not found when disk is not found in any zone, other error when its found itn multiple zones or error getting disk

Author: davidz627

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

@@ -75,7 +75,7 @@ func (cloud *FakeCloudProvider) RepairUnderspecifiedVolumeKey(ctx context.Contex
 				return volumeKey, nil
 			}
 		}
-		return nil, fmt.Errorf("Couldn't repair unspecified volume information")
+		return nil, notFoundError()
 	case meta.Regional:
 		if volumeKey.Region != common.UnspecifiedValue {
 			return volumeKey, nil

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

@@ -72,6 +72,7 @@ func (cloud *CloudProvider) RepairUnderspecifiedVolumeKey(ctx context.Context, v
 	}
 	switch volumeKey.Type() {
 	case meta.Zonal:
+		foundZone := ""
 		if volumeKey.Zone == common.UnspecifiedValue {
 			// list all zones, try to get disk in each zone
 			zones, err := cloud.ListZones(ctx, region)
@@ -80,13 +81,27 @@ func (cloud *CloudProvider) RepairUnderspecifiedVolumeKey(ctx context.Context, v
 			}
 			for _, zone := range zones {
 				_, err := cloud.getZonalDiskOrError(ctx, zone, volumeKey.Name)
-				if err == nil {
-					// If there is no error we have found a disk
-					volumeKey.Zone = zone
-					return volumeKey, nil
+				if err != nil {
+					if IsGCENotFoundError(err) {
+						// Couldn't find the disk in this zone so we keep
+						// looking
+						continue
+					}
+					// There is some miscellaneous error getting disk from zone
+					// so we return error immediately
+					return nil, err
 				}
+				if len(foundZone) > 0 {
+					return nil, fmt.Errorf("found disk %s in more than one zone: %s and %s", volumeKey.Name, foundZone, zone)
+				}
+				foundZone = zone
+			}
+
+			if len(foundZone) == 0 {
+				return nil, notFoundError()
 			}
-			return nil, fmt.Errorf("volume zone unspecified and unable to find in any of these zones %v", zones)
+			volumeKey.Zone = foundZone
+			return volumeKey, nil
 		}
 		return volumeKey, nil
 	case meta.Regional:

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

@@ -226,14 +226,17 @@ func (gceCS *GCEControllerServer) DeleteVolume(ctx context.Context, req *csi.Del
 	if err != nil {
 		// Cannot find volume associated with this ID because VolumeID is not in
 		// correct format, this is a success according to the Spec
-		klog.Warningf("Treating volume as deleted because volume id %s is invalid: %v", volumeID, err)
+		klog.Warningf("DeleteVolume treating volume as deleted because volume id %s is invalid: %v", volumeID, err)
 		return &csi.DeleteVolumeResponse{}, nil
 	}
 
 	volKey, err = gceCS.CloudProvider.RepairUnderspecifiedVolumeKey(ctx, volKey)
 	if err != nil {
-		klog.Warningf("Treating volume as deleted because cannot find volume %v: %v", volumeID, err)
-		return &csi.DeleteVolumeResponse{}, nil
+		if gce.IsGCENotFoundError(err) {
+			klog.Warningf("DeleteVolume treating volume as deleted because cannot find volume %v: %v", volumeID, err)
+			return &csi.DeleteVolumeResponse{}, nil
+		}
+		return nil, status.Errorf(codes.Internal, "DeleteVolume error repairing underspecified volume key: %v", err)
 	}
 
 	if acquired := gceCS.volumeLocks.TryAcquire(volumeID); !acquired {
@@ -274,7 +277,10 @@ func (gceCS *GCEControllerServer) ControllerPublishVolume(ctx context.Context, r
 
 	volKey, err = gceCS.CloudProvider.RepairUnderspecifiedVolumeKey(ctx, volKey)
 	if err != nil {
-		return nil, status.Error(codes.NotFound, fmt.Sprintf("Could not find volume with ID %v: %v", volumeID, err))
+		if gce.IsGCENotFoundError(err) {
+			return nil, status.Errorf(codes.NotFound, "ControllerPublishVolume could not find volume with ID %v: %v", volumeID, err)
+		}
+		return nil, status.Errorf(codes.Internal, "ControllerPublishVolume error repairing underspecified volume key: %v", err)
 	}
 
 	// Acquires the lock for the volume on that node only, because we need to support the ability