Add fsFreeze field to VolumeSnapshot

app/cmd/start.go

@@ -24,11 +24,13 @@ import (
 	healthpb "google.golang.org/grpc/health/grpc_health_v1"
 	"google.golang.org/grpc/keepalive"
 	"google.golang.org/grpc/reflection"
+	"k8s.io/mount-utils"
 
 	commonTypes "github.com/longhorn/go-common-libs/types"
 	helpernvme "github.com/longhorn/go-spdk-helper/pkg/nvme"
 	helpertypes "github.com/longhorn/go-spdk-helper/pkg/types"
 	helperutil "github.com/longhorn/go-spdk-helper/pkg/util"
+	engineutil "github.com/longhorn/longhorn-engine/pkg/util"
 	spdk "github.com/longhorn/longhorn-spdk-engine/pkg/spdk"
 	spdkutil "github.com/longhorn/longhorn-spdk-engine/pkg/util"
 	rpc "github.com/longhorn/types/pkg/generated/imrpc"
@@ -159,6 +161,34 @@ func cleanupStaledNvmeAndDmDevices() error {
 	return nil
 }
 
+func unfreezeFilesystems() error {
+	// We do not need to switch to the host mount namespace to get mount points here. Usually, longhorn-engine runs in a
+	// container that has / bind mounted to /host with at least HostToContainer (rslave) propagation.
+	// - If it does not, we likely can't do a namespace swap anyway, since we don't have access to /host/proc.
+	// - If it does, we just need to know where in the container we can access the mount points to unfreeze the file
+	//   system.
+	mounter := mount.New("")
+	mountPoints, err := mounter.List()
+	if err != nil {
+		return errors.Wrap(err, "failed to list mount points while starting up")
+	}
+
+	for _, mountPoint := range mountPoints {
+		if strings.Contains(mountPoint.Device, engineutil.DevicePathPrefix) {
+			// We do not actually expect any filesystems to be frozen. This is a best effort attempt to unfreeze them
+			// if somehow instance manager crashed at the wrong moment during a snapshot.
+			unfroze, err := engineutil.UnfreezeFilesystem(mountPoint.Path, nil)
+			if err != nil {
+				logrus.WithError(err).Warnf("Failed to unfreeze filesystem mounted at %v", mountPoint)
+			}
+			if unfroze {
+				logrus.Warnf("Unfroze filesystem mounted at %v", mountPoint)
+			}
+		}
+	}
+	return nil
+}
+
 func start(c *cli.Context) (err error) {
 	listen := c.String("listen")
 	logsDir := c.String("logs-dir")
@@ -183,6 +213,10 @@ func start(c *cli.Context) (err error) {
 		if err := cleanupStaledNvmeAndDmDevices(); err != nil {
 			return err
 		}
+	} else {
+		if err := unfreezeFilesystems(); err != nil {
+			return err
+		}
 	}
 
 	// setup tls config

go.mod

@@ -8,7 +8,7 @@ require (
 	github.com/longhorn/backupstore v0.0.0-20240509144945-3bce6e69af15
 	github.com/longhorn/go-common-libs v0.0.0-20240514074907-351459694cbf
 	github.com/longhorn/go-spdk-helper v0.0.0-20240514082311-4069f4804017
-	github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240509154612-5fdc92a2d526
+	github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240514224711-e39b7f0313b2
 	github.com/longhorn/longhorn-spdk-engine v0.0.0-20240516014845-759dfa872eae
 	github.com/longhorn/types v0.0.0-20240510221052-ab949bbedea3
 	github.com/pkg/errors v0.9.1
@@ -43,6 +43,7 @@ require (
 	github.com/jmespath/go-jmespath v0.4.0 // indirect
 	github.com/kr/pretty v0.3.1 // indirect
 	github.com/kr/text v0.2.0 // indirect
+	github.com/longhorn/go-iscsi-helper v0.0.0-20240427164656-e9439c0018ce // indirect
 	github.com/longhorn/nsfilelock v0.0.0-20200723175406-fa7c83ad0003 // indirect
 	github.com/longhorn/sparse-tools v0.0.0-20240427164751-a7b9f1b2c8a8 // indirect
 	github.com/matttproud/golang_protobuf_extensions/v2 v2.0.0 // indirect

go.sum

@@ -69,10 +69,12 @@ github.com/longhorn/backupstore v0.0.0-20240509144945-3bce6e69af15 h1:D838/RWPKm
 github.com/longhorn/backupstore v0.0.0-20240509144945-3bce6e69af15/go.mod h1:n210xpMUVrSn/W4Za/9BZhyXLCTVfAOq5lNdLrRSyz8=
 github.com/longhorn/go-common-libs v0.0.0-20240514074907-351459694cbf h1:3mjybe5dP3uJF2daMVv/U/IGNcp1dsrzHl4b6w8icPU=
 github.com/longhorn/go-common-libs v0.0.0-20240514074907-351459694cbf/go.mod h1:gFXUEciTv/03ncyA8CNrfVkbikvSWNqCYwwsTC3lFGg=
+github.com/longhorn/go-iscsi-helper v0.0.0-20240427164656-e9439c0018ce h1:PxKniE9F6IZ2DMKfmxDsbqeAxQI1TZhnw7/HOBMs1Is=
+github.com/longhorn/go-iscsi-helper v0.0.0-20240427164656-e9439c0018ce/go.mod h1:d9t3gtE+UPjescbCFluXd4xBc8OQT/JrC2cdkk2IXWQ=
 github.com/longhorn/go-spdk-helper v0.0.0-20240514082311-4069f4804017 h1:Cl8mEO7uP2T9ZDbqV3BK2xhrBVyjelY0WV9LDGdqb/Y=
 github.com/longhorn/go-spdk-helper v0.0.0-20240514082311-4069f4804017/go.mod h1:SgKStbsHMJqVjf/uDpS7KYQakOp2jHBRSbRnMk6iNqo=
-github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240509154612-5fdc92a2d526 h1:JvSWv3PBV89p6TsM2DQhkm/FB4L95U9l2nrPomzT9KA=
-github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240509154612-5fdc92a2d526/go.mod h1:SvLrPprzk/iXFG3G4wDbWXD1kePlxg12Oui96FLBHAg=
+github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240514224711-e39b7f0313b2 h1:mG7wUWZrJ5arEZWaghm73ywH09LUZo77kGUaHN7nzyg=
+github.com/longhorn/longhorn-engine v1.7.0-dev.0.20240514224711-e39b7f0313b2/go.mod h1:G+iGNlJm1RKuKXqz4Igh0b5miiRblld2C9GL9hs4ROk=
 github.com/longhorn/longhorn-spdk-engine v0.0.0-20240516014845-759dfa872eae h1:07uvZWinlp6Zwvuhacv9Lv3NzRM46UKPu0xIAXP7VEA=
 github.com/longhorn/longhorn-spdk-engine v0.0.0-20240516014845-759dfa872eae/go.mod h1:i8arwLcEawuN9rdHmo3doVhg9fQFPh0q+GJaKR2FOxk=
 github.com/longhorn/nsfilelock v0.0.0-20200723175406-fa7c83ad0003 h1:Jw9uANsGcHTxp6HcC++/vN17LfeuDmozHI2j6DoZf5E=

pkg/client/proxy_snapshot.go

@@ -12,7 +12,7 @@ import (
 )
 
 func (c *ProxyClient) VolumeSnapshot(dataEngine, engineName, volumeName, serviceAddress,
-	volumeSnapshotName string, labels map[string]string) (snapshotName string, err error) {
+	volumeSnapshotName string, labels map[string]string, freezeFilesystem bool) (snapshotName string, err error) {
 	input := map[string]string{
 		"engineName":     engineName,
 		"volumeName":     volumeName,
@@ -55,8 +55,9 @@ func (c *ProxyClient) VolumeSnapshot(dataEngine, engineName, volumeName, service
 			VolumeName:         volumeName,
 		},
 		SnapshotVolume: &enginerpc.VolumeSnapshotRequest{
-			Name:   volumeSnapshotName,
-			Labels: labels,
+			Name:             volumeSnapshotName,
+			Labels:           labels,
+			FreezeFilesystem: freezeFilesystem,
 		},
 	}
 	recv, err := c.service.VolumeSnapshot(getContextWithGRPCTimeout(c.ctx), req)

pkg/proxy/snapshot.go

@@ -41,7 +41,7 @@ func (ops V1DataEngineProxyOps) VolumeSnapshot(ctx context.Context, req *rpc.Eng
 	}
 	defer c.Close()
 
-	recv, err := c.VolumeSnapshot(req.SnapshotVolume.Name, req.SnapshotVolume.Labels)
+	recv, err := c.VolumeSnapshot(req.SnapshotVolume.Name, req.SnapshotVolume.Labels, req.SnapshotVolume.FreezeFilesystem)
 	if err != nil {
 		return nil, err
 	}