fix: improve handling of restore operations

- restore operations are split into a new flow
 - added support displaying restore operation percentage and other
   details in tree view

internal/api/backresthandler.go

@@ -400,12 +400,7 @@ func (s *BackrestHandler) Restore(ctx context.Context, req *connect.Request[v1.R
 	}
 
 	at := time.Now()
-
-	flowID, err := tasks.FlowIDForSnapshotID(s.oplog, req.Msg.SnapshotId)
-	if err != nil {
-		return nil, fmt.Errorf("failed to get flow ID for snapshot %q: %w", req.Msg.SnapshotId, err)
-	}
-	s.orchestrator.ScheduleTask(tasks.NewOneoffRestoreTask(req.Msg.RepoId, req.Msg.PlanId, flowID, at, req.Msg.SnapshotId, req.Msg.Path, req.Msg.Target), tasks.TaskPriorityInteractive+tasks.TaskPriorityDefault)
+	s.orchestrator.ScheduleTask(tasks.NewOneoffRestoreTask(req.Msg.RepoId, req.Msg.PlanId, 0 /* flowID */, at, req.Msg.SnapshotId, req.Msg.Path, req.Msg.Target), tasks.TaskPriorityInteractive+tasks.TaskPriorityDefault)
 
 	return connect.NewResponse(&emptypb.Empty{}), nil
 }

internal/orchestrator/tasks/taskcollectgarbage.go

@@ -13,16 +13,20 @@ import (
 const (
 	gcStartupDelay = 60 * time.Second
 	gcInterval     = 24 * time.Hour
-	// keep operations that are eligible for gc for 30 days OR up to a limit of 100 for any one plan.
-	// an operation is eligible for gc if:
-	// - it has no snapshot associated with it
-	// - it has a forgotten snapshot associated with it
-	gcHistoryAge      = 30 * 24 * time.Hour
-	gcHistoryMaxCount = 1000
-	// keep stats operations for 1 year (they're small and useful for long term trends)
-	gcHistoryStatsAge = 365 * 24 * time.Hour
 )
 
+// gcAgeForOperation returns the age at which an operation is eligible for garbage collection.
+func gcAgeForOperation(op *v1.Operation) time.Duration {
+	switch op.Op.(type) {
+	// stats, check, and prune operations are kept for a year
+	case *v1.Operation_OperationStats, *v1.Operation_OperationCheck, *v1.Operation_OperationPrune:
+		return 365 * 24 * time.Hour
+	// all other operations are kept for 30 days
+	default:
+		return 30 * 24 * time.Hour
+	}
+}
+
 type CollectGarbageTask struct {
 	BaseTask
 	firstRun bool
@@ -83,11 +87,8 @@ func (t *CollectGarbageTask) gcOperations(oplog *oplog.OpLog) error {
 		forgot, ok := snapshotForgottenForFlow[op.FlowId]
 		if !ok {
 			// no snapshot associated with this flow; check if it's old enough to be gc'd
-			maxAgeForType := gcHistoryAge.Milliseconds()
-			if _, isStats := op.Op.(*v1.Operation_OperationStats); isStats {
-				maxAgeForType = gcHistoryStatsAge.Milliseconds()
-			}
-			if curTime-op.UnixTimeStartMs > maxAgeForType {
+			maxAgeForOperation := gcAgeForOperation(op)
+			if curTime-op.UnixTimeStartMs > maxAgeForOperation.Milliseconds() {
 				forgetIDs = append(forgetIDs, op.Id)
 			}
 		} else if forgot {
@@ -107,17 +108,3 @@ func (t *CollectGarbageTask) gcOperations(oplog *oplog.OpLog) error {
 		zap.Any("operations_removed", len(forgetIDs)))
 	return nil
 }
-
-func (t *CollectGarbageTask) Cancel(withStatus v1.OperationStatus) error {
-	return nil
-}
-
-func (t *CollectGarbageTask) OperationId() int64 {
-	return 0
-}
-
-type gcOpInfo struct {
-	id        int64 // operation ID
-	timestamp int64 // unix time milliseconds
-	isStats   bool  // true if this is a stats operation
-}

internal/orchestrator/tasks/taskrestore.go

@@ -77,7 +77,7 @@ func restoreHelper(ctx context.Context, st ScheduledTask, taskRunner TaskRunner,
 
 		zap.S().Infof("restore progress: %v", entry)
 
-		restoreOp.Status = entry
+		restoreOp.LastStatus = entry
 
 		sendWg.Add(1)
 		go func() {
@@ -91,7 +91,7 @@ func restoreHelper(ctx context.Context, st ScheduledTask, taskRunner TaskRunner,
 	if err != nil {
 		return fmt.Errorf("restore failed: %w", err)
 	}
-	restoreOp.Status = summary
+	restoreOp.LastStatus = summary
 
 	return nil
 }