fix: pre-clean snapshot dir before store

internal/advancer/advancer.go

@@ -465,6 +465,13 @@ func (s *Service) createSnapshot(ctx context.Context, app *Application, machine
 		"input", input.Index,
 		"path", snapshotPath)
 
+	// Guard against partial store by cleaning up before a retry
+	if _, err := os.Stat(snapshotPath); err == nil {
+		if err := s.removeSnapshot(snapshotPath, app.Name); err != nil {
+			return fmt.Errorf("failed to clean stale snapshot directory: %w", err)
+		}
+	}
+
 	// Ensure the parent directory exists
 	if err := os.MkdirAll(s.snapshotsDir, 0755); err != nil { //nolint: mnd
 		return fmt.Errorf("failed to create snapshots directory: %w", err)

internal/advancer/advancer_test.go

@@ -9,6 +9,7 @@ import (
 	"encoding/json"
 	"errors"
 	"fmt"
+	"io/fs"
 	mrand "math/rand"
 	"os"
 	"path/filepath"
@@ -937,6 +938,27 @@ func (s *AdvancerSuite) TestCreateSnapshot() {
 		require.True(os.IsNotExist(statErr))
 	})
 
+	s.Run("OverwritesPartialPriorAttempt", func() {
+
+		require := s.Require()
+		env, machine, tmpDir := setupCreateSnapshot()
+
+		prevPath := filepath.Join(tmpDir, "testapp_epoch0_input0")
+		require.Nil(os.MkdirAll(prevPath, 0755))
+
+		input := repotest.NewInputBuilder().WithIndex(0).WithEpochIndex(0).Build()
+		input.EpochApplicationID = env.app.Application.ID
+
+		err := env.service.createSnapshot(context.Background(), env.app.Application, machine, input)
+		require.Nil(err)
+		require.NotNil(input.SnapshotURI)
+		require.Equal(prevPath, *input.SnapshotURI)
+
+		// This works because the machine mock doesn't create directories
+		_, statErr := os.Stat(prevPath)
+		require.True(errors.Is(statErr, fs.ErrNotExist))
+	})
+
 	s.Run("CreateSnapshotError", func() {
 		require := s.Require()
 		env, machine, _ := setupCreateSnapshot()