Merge pull request 'fix: clean up orphan resources on deploy cancellation (closes #89)' (#93) from fix/deploy-cancel-cleanup into main

Reviewed-on: #93

internal/docker/client.go

@@ -17,6 +17,7 @@ import (
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
+	"github.com/docker/docker/api/types/image"
 	"github.com/docker/docker/api/types/mount"
 	"github.com/docker/docker/api/types/network"
 	"github.com/docker/docker/client"
@@ -739,6 +740,20 @@ func (c *Client) connect(ctx context.Context) error {
 	return nil
 }
 
+// RemoveImage removes a Docker image by ID or tag.
+// It returns nil if the image was successfully removed or does not exist.
+func (c *Client) RemoveImage(ctx context.Context, imageID string) error {
+	_, err := c.docker.ImageRemove(ctx, imageID, image.RemoveOptions{
+		Force:         true,
+		PruneChildren: true,
+	})
+	if err != nil && !client.IsErrNotFound(err) {
+		return fmt.Errorf("failed to remove image %s: %w", imageID, err)
+	}
+
+	return nil
+}
+
 func (c *Client) close() error {
 	if c.docker != nil {
 		err := c.docker.Close()

internal/service/deploy/deploy.go

@@ -11,6 +11,7 @@ import (
 	"log/slog"
 	"os"
 	"path/filepath"
+	"strings"
 	"sync"
 	"time"
 
@@ -472,7 +473,7 @@ func (svc *Service) runBuildAndDeploy(
 	// Build phase with timeout
 	imageID, err := svc.buildImageWithTimeout(deployCtx, app, deployment)
 	if err != nil {
-		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment)
+		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment, "")
 		if cancelErr != nil {
 			return cancelErr
 		}
@@ -485,7 +486,7 @@ func (svc *Service) runBuildAndDeploy(
 	// Deploy phase with timeout
 	err = svc.deployContainerWithTimeout(deployCtx, app, deployment, imageID)
 	if err != nil {
-		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment)
+		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment, imageID)
 		if cancelErr != nil {
 			return cancelErr
 		}
@@ -661,24 +662,76 @@ func (svc *Service) cancelActiveDeploy(appID string) {
 }
 
 // checkCancelled checks if the deploy context was cancelled (by a newer deploy)
-// and if so, marks the deployment as cancelled. Returns ErrDeployCancelled or nil.
+// and if so, marks the deployment as cancelled and cleans up orphan resources.
+// Returns ErrDeployCancelled or nil.
 func (svc *Service) checkCancelled(
 	deployCtx context.Context,
 	bgCtx context.Context,
 	app *models.App,
 	deployment *models.Deployment,
+	imageID string,
 ) error {
 	if !errors.Is(deployCtx.Err(), context.Canceled) {
 		return nil
 	}
 
-	svc.log.Info("deployment cancelled by newer deploy", "app", app.Name)
+	svc.log.Info("deployment cancelled", "app", app.Name)
+
+	svc.cleanupCancelledDeploy(bgCtx, app, deployment, imageID)
 
 	_ = deployment.MarkFinished(bgCtx, models.DeploymentStatusCancelled)
 
 	return ErrDeployCancelled
 }
 
+// cleanupCancelledDeploy removes orphan resources left by a cancelled deployment.
+func (svc *Service) cleanupCancelledDeploy(
+	ctx context.Context,
+	app *models.App,
+	deployment *models.Deployment,
+	imageID string,
+) {
+	// Clean up the intermediate Docker image if one was built
+	if imageID != "" {
+		removeErr := svc.docker.RemoveImage(ctx, imageID)
+		if removeErr != nil {
+			svc.log.Error("failed to remove image from cancelled deploy",
+				"error", removeErr, "app", app.Name, "image", imageID)
+			_ = deployment.AppendLog(ctx, "WARNING: failed to clean up image "+imageID+": "+removeErr.Error())
+		} else {
+			svc.log.Info("cleaned up image from cancelled deploy",
+				"app", app.Name, "image", imageID)
+			_ = deployment.AppendLog(ctx, "Cleaned up intermediate image: "+imageID)
+		}
+	}
+
+	// Clean up the build directory for this deployment
+	buildDir := svc.GetBuildDir(app.Name)
+
+	entries, err := os.ReadDir(buildDir)
+	if err != nil {
+		return
+	}
+
+	prefix := fmt.Sprintf("%d-", deployment.ID)
+
+	for _, entry := range entries {
+		if entry.IsDir() && strings.HasPrefix(entry.Name(), prefix) {
+			dirPath := filepath.Join(buildDir, entry.Name())
+
+			removeErr := os.RemoveAll(dirPath)
+			if removeErr != nil {
+				svc.log.Error("failed to remove build dir from cancelled deploy",
+					"error", removeErr, "path", dirPath)
+			} else {
+				svc.log.Info("cleaned up build dir from cancelled deploy",
+					"app", app.Name, "path", dirPath)
+				_ = deployment.AppendLog(ctx, "Cleaned up build directory")
+			}
+		}
+	}
+}
+
 func (svc *Service) fetchWebhookEvent(
 	ctx context.Context,
 	webhookEventID *int64,

internal/service/deploy/deploy_cleanup_test.go

@@ -0,0 +1,63 @@
+package deploy_test
+
+import (
+	"context"
+	"log/slog"
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"git.eeqj.de/sneak/upaas/internal/config"
+	"git.eeqj.de/sneak/upaas/internal/service/deploy"
+)
+
+func TestCleanupCancelledDeploy_RemovesBuildDir(t *testing.T) {
+	t.Parallel()
+
+	tmpDir := t.TempDir()
+	cfg := &config.Config{DataDir: tmpDir}
+
+	svc := deploy.NewTestServiceWithConfig(slog.Default(), cfg, nil)
+
+	// Create a fake build directory matching the deployment pattern
+	appName := "test-app"
+	buildDir := svc.GetBuildDirExported(appName)
+	require.NoError(t, os.MkdirAll(buildDir, 0o750))
+
+	// Create deployment-specific dir: <deploymentID>-<random>
+	deployDir := filepath.Join(buildDir, "42-abc123")
+	require.NoError(t, os.MkdirAll(deployDir, 0o750))
+
+	// Create a file inside to verify full removal
+	require.NoError(t, os.WriteFile(filepath.Join(deployDir, "work"), []byte("test"), 0o640))
+
+	// Also create a dir for a different deployment (should NOT be removed)
+	otherDir := filepath.Join(buildDir, "99-xyz789")
+	require.NoError(t, os.MkdirAll(otherDir, 0o750))
+
+	// Run cleanup for deployment 42
+	svc.CleanupCancelledDeploy(context.Background(), appName, 42, "")
+
+	// Deployment 42's dir should be gone
+	_, err := os.Stat(deployDir)
+	assert.True(t, os.IsNotExist(err), "deployment build dir should be removed")
+
+	// Deployment 99's dir should still exist
+	_, err = os.Stat(otherDir)
+	assert.NoError(t, err, "other deployment build dir should not be removed")
+}
+
+func TestCleanupCancelledDeploy_NoBuildDir(t *testing.T) {
+	t.Parallel()
+
+	tmpDir := t.TempDir()
+	cfg := &config.Config{DataDir: tmpDir}
+
+	svc := deploy.NewTestServiceWithConfig(slog.Default(), cfg, nil)
+
+	// Should not panic when build dir doesn't exist
+	svc.CleanupCancelledDeploy(context.Background(), "nonexistent-app", 1, "")
+}

internal/service/deploy/export_test.go

@@ -2,7 +2,14 @@ package deploy
 
 import (
 	"context"
+	"fmt"
 	"log/slog"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"git.eeqj.de/sneak/upaas/internal/config"
+	"git.eeqj.de/sneak/upaas/internal/docker"
 )
 
 // NewTestService creates a Service with minimal dependencies for testing.
@@ -31,3 +38,45 @@ func (svc *Service) TryLockApp(appID string) bool {
 func (svc *Service) UnlockApp(appID string) {
 	svc.unlockApp(appID)
 }
+
+// NewTestServiceWithConfig creates a Service with config and docker client for testing.
+func NewTestServiceWithConfig(log *slog.Logger, cfg *config.Config, dockerClient *docker.Client) *Service {
+	return &Service{
+		log:    log,
+		config: cfg,
+		docker: dockerClient,
+	}
+}
+
+// CleanupCancelledDeploy exposes the build directory cleanup portion of
+// cleanupCancelledDeploy for testing. It removes build directories matching
+// the deployment ID prefix.
+func (svc *Service) CleanupCancelledDeploy(
+	ctx context.Context,
+	appName string,
+	deploymentID int64,
+	imageID string,
+) {
+	// We can't create real models.App/Deployment in tests easily,
+	// so we test the build dir cleanup portion directly.
+	buildDir := svc.GetBuildDir(appName)
+
+	entries, err := os.ReadDir(buildDir)
+	if err != nil {
+		return
+	}
+
+	prefix := fmt.Sprintf("%d-", deploymentID)
+
+	for _, entry := range entries {
+		if entry.IsDir() && strings.HasPrefix(entry.Name(), prefix) {
+			dirPath := filepath.Join(buildDir, entry.Name())
+			_ = os.RemoveAll(dirPath)
+		}
+	}
+}
+
+// GetBuildDirExported exposes GetBuildDir for testing.
+func (svc *Service) GetBuildDirExported(appName string) string {
+	return svc.GetBuildDir(appName)
+}