rbd: do not execute rbd sparsify when volume is in use (backport #3985)

internal/csi-addons/rbd/reclaimspace.go

@@ -18,11 +18,13 @@ package rbd
 
 import (
 	"context"
+	"errors"
 	"fmt"
 
 	csicommon "github.com/ceph/ceph-csi/internal/csi-common"
 	rbdutil "github.com/ceph/ceph-csi/internal/rbd"
 	"github.com/ceph/ceph-csi/internal/util"
+	"github.com/ceph/ceph-csi/internal/util/log"
 
 	"github.com/container-storage-interface/spec/lib/go/csi"
 	rs "github.com/csi-addons/spec/lib/go/reclaimspace"
@@ -69,6 +71,13 @@ func (rscs *ReclaimSpaceControllerServer) ControllerReclaimSpace(
 	defer rbdVol.Destroy()
 
 	err = rbdVol.Sparsify()
+	if errors.Is(err, rbdutil.ErrImageInUse) {
+		// FIXME: https://github.com/csi-addons/kubernetes-csi-addons/issues/406.
+		// treat sparsify call as no-op if volume is in use.
+		log.DebugLog(ctx, fmt.Sprintf("volume with ID %q is in use, skipping sparsify operation", volumeID))
+
+		return &rs.ControllerReclaimSpaceResponse{}, nil
+	}
 	if err != nil {
 		// TODO: check for different error codes?
 		return nil, status.Errorf(codes.Internal, "failed to sparsify volume %q: %s", rbdVol, err.Error())

internal/rbd/diskusage.go

@@ -23,7 +23,18 @@ import (
 // Sparsify checks the size of the objects in the RBD image and calls
 // rbd_sparify() to free zero-filled blocks and reduce the storage consumption
 // of the image.
+// This function will return ErrImageInUse if the image is in use, since
+// sparsifying an image on which i/o is in progress is not optimal.
 func (ri *rbdImage) Sparsify() error {
+	inUse, err := ri.isInUse()
+	if err != nil {
+		return fmt.Errorf("failed to check if image is in use: %w", err)
+	}
+	if inUse {
+		// if the image is in use, we should not sparsify it, return ErrImageInUse.
+		return ErrImageInUse
+	}
+
 	image, err := ri.open()
 	if err != nil {
 		return err

internal/rbd/errors.go

@@ -45,4 +45,6 @@ var (
 	// ErrLastSyncTimeNotFound is returned when last sync time is not found for
 	// the image.
 	ErrLastSyncTimeNotFound = errors.New("last sync time not found")
+	// ErrImageInUse is returned when the image is in use.
+	ErrImageInUse = errors.New("image is in use")
 )