pillar: add immediate snapshot feature

this feature allows to make an "immediate" snapshot,
i.e. the request triggers a shutdown of the edge app,
then does the snapshot and brings the edge app back
to running state

Signed-off-by: Christoph Ostarek <[email protected]>

pkg/pillar/cmd/zedagent/reportinfo.go

@@ -648,7 +648,7 @@ func PublishDeviceInfoToZedCloud(ctx *zedagentContext, dest destinationBitset) {
 	// device returns a runtime error. Similarly, we only support enforced application network
 	// interface order for the KVM hypervisor. If enabled for application deployed under Xen
 	// or Kubevirt hypervisor, EVE returns error and the application will not be started.
-	ReportDeviceInfo.ApiCapability = info.APICapability_API_CAPABILITY_NTPS_FQDN
+	ReportDeviceInfo.ApiCapability = info.APICapability_API_CAPABILITY_VOLUME_SNAPSHOTS_IMMEDIATE
 
 	// Report if there is a local override of profile
 	if ctx.getconfigCtx.sideController.currentProfile !=

pkg/pillar/cmd/zedmanager/updatestatus.go

@@ -86,7 +86,7 @@ func removeAIStatus(ctx *zedmanagerContext, status *types.AppInstanceStatus) {
 	// The VM has been just shutdown in a result of the purge&update command coming from the controller.
 	if !uninstall && domainStatus != nil && !domainStatus.Activated {
 		// We should do it before the doRemove is called, so that all the volumes are still available.
-		if status.SnapStatus.SnapshotOnUpgrade && len(status.SnapStatus.PreparedVolumesSnapshotConfigs) > 0 {
+		if status.SnapStatus.SnapshotTakenType != types.NoSnapshotTake && len(status.SnapStatus.PreparedVolumesSnapshotConfigs) > 0 {
 			// Check whether there are snapshots to be deleted first (not to exceed the maximum number of snapshots).
 			if len(status.SnapStatus.SnapshotsToBeDeleted) > 0 {
 				triggerSnapshotDeletion(status.SnapStatus.SnapshotsToBeDeleted, ctx, status)

pkg/pillar/cmd/zedmanager/zedmanager.go

@@ -829,18 +829,26 @@ func lookupLocalAppInstanceConfig(ctx *zedmanagerContext, key string) *types.App
 	return &config
 }
 
-func isSnapshotRequestedOnUpdate(status *types.AppInstanceStatus, config types.AppInstanceConfig) bool {
-	if config.Snapshot.Snapshots == nil {
-		return false
-	}
+func snapshotRequestedType(status *types.AppInstanceStatus, config types.AppInstanceConfig) types.SnapshotTakenTime {
+	ret := types.NoSnapshotTake
+
 	for _, snap := range config.Snapshot.Snapshots {
-		// VM should be marked to be snapshotted on update only if there is any snapshot request of the type "on update"
-		// that is not handled yet.
-		if snap.SnapshotType == types.SnapshotTypeAppUpdate && lookupAvailableSnapshot(status, snap.SnapshotID) == nil {
-			return true
+		if lookupAvailableSnapshot(status, snap.SnapshotID) != nil {
+			continue
+		}
+
+		switch snap.SnapshotType {
+		case types.SnapshotTypeUnspecified:
+			ret = types.NoSnapshotTake
+		case types.SnapshotTypeAppUpdate:
+			ret = types.SnapshotOnUpgrade
+		case types.SnapshotTypeImmediate:
+			// Immediate snapshots have precedence over snapshots on upgrade
+			return types.SnapshotImmediate
 		}
 	}
-	return false
+
+	return ret
 }
 
 // removeNUnpublishedSnapshotRequests removes up to n snapshot requests that have not been triggered yet
@@ -985,7 +993,7 @@ func isNewSnapshotRequest(id string, status *types.AppInstanceStatus) bool {
 
 // Update the snapshot related fields in the AppInstanceStatus
 func updateSnapshotsInAIStatus(status *types.AppInstanceStatus, config types.AppInstanceConfig) {
-	status.SnapStatus.SnapshotOnUpgrade = isSnapshotRequestedOnUpdate(status, config)
+	status.SnapStatus.SnapshotTakenType = snapshotRequestedType(status, config)
 	//markReportedSnapshots(status, config)
 	status.SnapStatus.MaxSnapshots = config.Snapshot.MaxSnapshots
 	snapshotsToBeDeleted := getSnapshotsToBeDeleted(config, status)
@@ -1021,7 +1029,7 @@ func updateSnapshotsInAIStatus(status *types.AppInstanceStatus, config types.App
 func prepareVolumesSnapshotConfigs(ctx *zedmanagerContext, config types.AppInstanceConfig, status *types.AppInstanceStatus) []types.VolumesSnapshotConfig {
 	var volumesSnapshotConfigList []types.VolumesSnapshotConfig
 	for _, snapshot := range status.SnapStatus.RequestedSnapshots {
-		if snapshot.Snapshot.SnapshotType == types.SnapshotTypeAppUpdate {
+		if snapshot.Snapshot.SnapshotType == types.SnapshotTypeAppUpdate || snapshot.Snapshot.SnapshotType == types.SnapshotTypeImmediate {
 			log.Noticef("Creating volumesSnapshotConfig for snapshot %s", snapshot.Snapshot.SnapshotID)
 			volumesSnapshotConfig := types.VolumesSnapshotConfig{
 				SnapshotID: snapshot.Snapshot.SnapshotID,
@@ -1264,9 +1272,11 @@ func handleModify(ctxArg interface{}, key string,
 	if needPurge {
 		needRestart = false
 	}
+
 	// A snapshot is deemed necessary whenever the application requires a restart, as this typically
 	// indicates a significant change in the application, such as an upgrade.
-	if status.SnapStatus.SnapshotOnUpgrade && (needRestart || needPurge) {
+	if status.SnapStatus.SnapshotTakenType == types.SnapshotImmediate ||
+		(status.SnapStatus.SnapshotTakenType == types.SnapshotOnUpgrade && (needRestart || needPurge)) {
 		// Save the list of the volumes that need to be backed up. We will use this list to create the snapshot when
 		// it's triggered. We cannot trigger the snapshot creation here immediately, as the VM
 		// should be stopped first. But we still need to save the list of volumes that are known only at this point.
@@ -1279,6 +1289,17 @@ func handleModify(ctxArg interface{}, key string,
 		}
 	}
 
+	if status.SnapStatus.SnapshotTakenType == types.SnapshotImmediate {
+		// this is okay as a clash with updated volumes is prevented later by
+		// "Need purge due to %s but not a purgeCmd", so if the user
+		// updates the volumes but does not set the purgeCmd but does
+		// an immediate snapshot, the immediate snapshot will not trigger
+		// an update of the volumes as the purgeCmd is missing
+		status.PurgeInprogress = types.DownloadAndVerify
+		status.PurgeStartedAt = time.Now()
+		restartReason = "Restart to create immediate snapshot"
+	}
+
 	if config.RestartCmd.Counter != oldConfig.RestartCmd.Counter ||
 		config.LocalRestartCmd.Counter != oldConfig.LocalRestartCmd.Counter {
 

pkg/pillar/types/zedmanagertypes.go

@@ -27,10 +27,12 @@ type UUIDandVersion struct {
 type SnapshotType int32
 
 const (
-	// SnapshotTypeUnspecified is the default value, and should not be used in practice
+	// SnapshotTypeUnspecified is the default value, and means no snapshot should be taken
 	SnapshotTypeUnspecified SnapshotType = 0
 	// SnapshotTypeAppUpdate is used when the snapshot is created as a result of an app update
 	SnapshotTypeAppUpdate SnapshotType = 1
+	// SnapshotTypeImmediate is used when the snapshot is created immediately
+	SnapshotTypeImmediate SnapshotType = 2
 )
 
 func (s SnapshotType) String() string {
@@ -39,6 +41,8 @@ func (s SnapshotType) String() string {
 		return "SnapshotTypeUnspecified"
 	case SnapshotTypeAppUpdate:
 		return "SnapshotTypeAppUpdate"
+	case SnapshotTypeImmediate:
+		return "SnapshotTypeImmediate"
 	default:
 		return fmt.Sprintf("Unknown SnapshotType %d", s)
 	}
@@ -49,6 +53,8 @@ func (s SnapshotType) ConvertToInfoSnapshotType() info.SnapshotType {
 	switch s {
 	case SnapshotTypeAppUpdate:
 		return info.SnapshotType_SNAPSHOT_TYPE_APP_UPDATE
+	case SnapshotTypeImmediate:
+		return info.SnapshotType_SNAPSHOT_TYPE_IMMEDIATE
 	default:
 		return info.SnapshotType_SNAPSHOT_TYPE_UNSPECIFIED
 	}
@@ -217,6 +223,18 @@ func (config AppInstanceConfig) Key() string {
 	return config.UUIDandVersion.UUID.String()
 }
 
+// SnapshotTakenTime describes when a snapshot should be taken, see info below
+type SnapshotTakenTime uint8
+
+const (
+	// NoSnapshotTake indicated no snapshot should be taken, f.e. because it already exists.
+	NoSnapshotTake SnapshotTakenTime = iota
+	// SnapshotImmediate indicates whether a snapshot should be taken immediately.
+	SnapshotImmediate
+	// SnapshotOnUpgrade indicates whether a snapshot should be taken during the app instance update.
+	SnapshotOnUpgrade
+)
+
 // SnapshottingStatus contains the snapshot information for the app instance.
 type SnapshottingStatus struct {
 	// MaxSnapshots indicates the maximum number of snapshots to be kept for the app instance.
@@ -229,8 +247,8 @@ type SnapshottingStatus struct {
 	SnapshotsToBeDeleted []SnapshotDesc
 	// PreparedVolumesSnapshotConfigs contains the list of snapshots to be triggered for the app instance.
 	PreparedVolumesSnapshotConfigs []VolumesSnapshotConfig
-	// SnapshotOnUpgrade indicates whether a snapshot should be taken during the app instance update.
-	SnapshotOnUpgrade bool
+	// SnapshotTakenType indicates if and when a snapshot should be taken
+	SnapshotTakenType SnapshotTakenTime
 	// HasRollbackRequest indicates whether there are any rollback requests for the app instance.
 	// Set to true when a rollback is requested by controller, set to false when the rollback is triggered.
 	HasRollbackRequest bool