feat: add user-facing deployment cancel endpoint

Add POST /apps/{id}/deployments/cancel endpoint that allows users to
cancel in-progress deployments via the web UI.

Changes:
- Add CancelDeploy() and HasActiveDeploy() public methods to deploy service
- Add HandleCancelDeploy handler
- Wire route in routes.go
- Add cancel button to app detail template (shown during active deployments)
- Add handler tests for cancel endpoint

fixes #66

internal/handlers/app.go

@@ -335,7 +335,7 @@ func (h *Handlers) HandleAppDeploy() http.HandlerFunc {
 		deployCtx := context.WithoutCancel(request.Context())
 
 		go func(ctx context.Context, appToDeploy *models.App) {
-			deployErr := h.deploy.Deploy(ctx, appToDeploy, nil)
+			deployErr := h.deploy.Deploy(ctx, appToDeploy, nil, false)
 			if deployErr != nil {
 				h.log.Error(
 					"deployment failed",
@@ -354,6 +354,32 @@ func (h *Handlers) HandleAppDeploy() http.HandlerFunc {
 	}
 }
 
+// HandleCancelDeploy cancels an in-progress deployment for an app.
+func (h *Handlers) HandleCancelDeploy() http.HandlerFunc {
+	return func(writer http.ResponseWriter, request *http.Request) {
+		appID := chi.URLParam(request, "id")
+
+		application, findErr := models.FindApp(request.Context(), h.db, appID)
+		if findErr != nil || application == nil {
+			http.NotFound(writer, request)
+
+			return
+		}
+
+		cancelled := h.deploy.CancelDeploy(application.ID)
+		if cancelled {
+			h.log.Info("deployment cancelled by user", "app", application.Name)
+		}
+
+		http.Redirect(
+			writer,
+			request,
+			"/apps/"+application.ID,
+			http.StatusSeeOther,
+		)
+	}
+}
+
 // HandleAppDeployments returns the deployments history handler.
 func (h *Handlers) HandleAppDeployments() http.HandlerFunc {
 	tmpl := templates.GetParsed()

internal/handlers/handlers_test.go

@@ -684,6 +684,47 @@ func TestDeletePortOwnershipVerification(t *testing.T) {
 	assert.NotNil(t, found, "port should still exist after IDOR attempt")
 }
 
+func TestHandleCancelDeployRedirects(t *testing.T) {
+	t.Parallel()
+
+	testCtx := setupTestHandlers(t)
+
+	createdApp := createTestApp(t, testCtx, "cancel-deploy-app")
+
+	request := httptest.NewRequest(
+		http.MethodPost,
+		"/apps/"+createdApp.ID+"/deployments/cancel",
+		nil,
+	)
+	request = addChiURLParams(request, map[string]string{"id": createdApp.ID})
+	recorder := httptest.NewRecorder()
+
+	handler := testCtx.handlers.HandleCancelDeploy()
+	handler.ServeHTTP(recorder, request)
+
+	assert.Equal(t, http.StatusSeeOther, recorder.Code)
+	assert.Equal(t, "/apps/"+createdApp.ID, recorder.Header().Get("Location"))
+}
+
+func TestHandleCancelDeployReturns404ForUnknownApp(t *testing.T) {
+	t.Parallel()
+
+	testCtx := setupTestHandlers(t)
+
+	request := httptest.NewRequest(
+		http.MethodPost,
+		"/apps/nonexistent/deployments/cancel",
+		nil,
+	)
+	request = addChiURLParams(request, map[string]string{"id": "nonexistent"})
+	recorder := httptest.NewRecorder()
+
+	handler := testCtx.handlers.HandleCancelDeploy()
+	handler.ServeHTTP(recorder, request)
+
+	assert.Equal(t, http.StatusNotFound, recorder.Code)
+}
+
 func TestHandleWebhookReturns404ForUnknownSecret(t *testing.T) {
 	t.Parallel()
 

internal/models/deployment.go

@@ -19,6 +19,7 @@ const (
 	DeploymentStatusDeploying DeploymentStatus = "deploying"
 	DeploymentStatusSuccess   DeploymentStatus = "success"
 	DeploymentStatusFailed    DeploymentStatus = "failed"
+	DeploymentStatusCancelled DeploymentStatus = "cancelled"
 )
 
 // Display constants.

internal/server/routes.go

@@ -68,6 +68,7 @@ func (s *Server) SetupRoutes() {
 		r.Post("/apps/{id}", s.handlers.HandleAppUpdate())
 		r.Post("/apps/{id}/delete", s.handlers.HandleAppDelete())
 		r.Post("/apps/{id}/deploy", s.handlers.HandleAppDeploy())
+		r.Post("/apps/{id}/deployments/cancel", s.handlers.HandleCancelDeploy())
 		r.Get("/apps/{id}/deployments", s.handlers.HandleAppDeployments())
 		r.Get("/apps/{id}/deployments/{deploymentID}/logs", s.handlers.HandleDeploymentLogsAPI())
 		r.Get("/apps/{id}/deployments/{deploymentID}/download", s.handlers.HandleDeploymentLogDownload())

internal/service/deploy/deploy.go

@@ -43,6 +43,8 @@ var (
 	ErrContainerUnhealthy = errors.New("container unhealthy after 60 seconds")
 	// ErrDeploymentInProgress indicates another deployment is already running.
 	ErrDeploymentInProgress = errors.New("deployment already in progress for this app")
+	// ErrDeployCancelled indicates the deployment was cancelled by a newer deploy.
+	ErrDeployCancelled = errors.New("deployment cancelled by newer deploy")
 	// ErrBuildTimeout indicates the build phase exceeded the timeout.
 	ErrBuildTimeout = errors.New("build timeout exceeded")
 	// ErrDeployTimeout indicates the deploy phase exceeded the timeout.
@@ -205,15 +207,22 @@ type ServiceParams struct {
 	Notify   *notify.Service
 }
 
+// activeDeploy tracks a running deployment so it can be cancelled.
+type activeDeploy struct {
+	cancel context.CancelFunc
+	done   chan struct{}
+}
+
 // Service provides deployment functionality.
 type Service struct {
-	log      *slog.Logger
-	db       *database.Database
-	docker   *docker.Client
-	notify   *notify.Service
-	config   *config.Config
-	params   *ServiceParams
-	appLocks sync.Map // map[string]*sync.Mutex - per-app deployment locks
+	log           *slog.Logger
+	db            *database.Database
+	docker        *docker.Client
+	notify        *notify.Service
+	config        *config.Config
+	params        *ServiceParams
+	activeDeploys sync.Map // map[string]*activeDeploy - per-app active deployment tracking
+	appLocks      sync.Map // map[string]*sync.Mutex - per-app deployment locks
 }
 
 // New creates a new deploy Service.
@@ -274,12 +283,39 @@ func (svc *Service) GetLogFilePath(app *models.App, deployment *models.Deploymen
 	return filepath.Join(svc.config.DataDir, "logs", hostname, app.Name, filename)
 }
 
-// Deploy deploys an app.
+// HasActiveDeploy returns true if there is an active deployment for the given app.
+func (svc *Service) HasActiveDeploy(appID string) bool {
+	_, ok := svc.activeDeploys.Load(appID)
+
+	return ok
+}
+
+// CancelDeploy cancels any in-progress deployment for the given app
+// and waits for it to finish before returning. Returns true if a deployment
+// was cancelled, false if there was nothing to cancel.
+func (svc *Service) CancelDeploy(appID string) bool {
+	if !svc.HasActiveDeploy(appID) {
+		return false
+	}
+
+	svc.cancelActiveDeploy(appID)
+
+	return true
+}
+
+// Deploy deploys an app. If cancelExisting is true (e.g. webhook-triggered),
+// any in-progress deploy for the same app will be cancelled before starting.
+// If cancelExisting is false and a deploy is in progress, ErrDeploymentInProgress is returned.
 func (svc *Service) Deploy(
 	ctx context.Context,
 	app *models.App,
 	webhookEventID *int64,
+	cancelExisting bool,
 ) error {
+	if cancelExisting {
+		svc.cancelActiveDeploy(app.ID)
+	}
+
 	// Try to acquire per-app deployment lock
 	if !svc.tryLockApp(app.ID) {
 		svc.log.Warn("deployment already in progress", "app", app.Name)
@@ -288,45 +324,80 @@ func (svc *Service) Deploy(
 	}
 	defer svc.unlockApp(app.ID)
 
+	// Set up cancellable context and register as active deploy
+	deployCtx, cancel := context.WithCancel(ctx)
+	done := make(chan struct{})
+	ad := &activeDeploy{cancel: cancel, done: done}
+	svc.activeDeploys.Store(app.ID, ad)
+
+	defer func() {
+		cancel()
+		close(done)
+		svc.activeDeploys.Delete(app.ID)
+	}()
+
 	// Fetch webhook event and create deployment record
-	webhookEvent := svc.fetchWebhookEvent(ctx, webhookEventID)
+	webhookEvent := svc.fetchWebhookEvent(deployCtx, webhookEventID)
 
-	deployment, err := svc.createDeploymentRecord(ctx, app, webhookEventID, webhookEvent)
+	// Use a background context for DB operations that must complete regardless of cancellation
+	bgCtx := context.WithoutCancel(deployCtx)
+
+	deployment, err := svc.createDeploymentRecord(bgCtx, app, webhookEventID, webhookEvent)
 	if err != nil {
 		return err
 	}
 
-	svc.logWebhookPayload(ctx, deployment, webhookEvent)
+	svc.logWebhookPayload(bgCtx, deployment, webhookEvent)
 
-	err = svc.updateAppStatusBuilding(ctx, app)
+	err = svc.updateAppStatusBuilding(bgCtx, app)
 	if err != nil {
 		return err
 	}
 
-	svc.notify.NotifyBuildStart(ctx, app, deployment)
+	svc.notify.NotifyBuildStart(bgCtx, app, deployment)
 
+	return svc.runBuildAndDeploy(deployCtx, bgCtx, app, deployment)
+}
+
+// runBuildAndDeploy executes the build and deploy phases, handling cancellation.
+func (svc *Service) runBuildAndDeploy(
+	deployCtx context.Context,
+	bgCtx context.Context,
+	app *models.App,
+	deployment *models.Deployment,
+) error {
 	// Build phase with timeout
-	imageID, err := svc.buildImageWithTimeout(ctx, app, deployment)
+	imageID, err := svc.buildImageWithTimeout(deployCtx, app, deployment)
 	if err != nil {
+		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment)
+		if cancelErr != nil {
+			return cancelErr
+		}
+
 		return err
 	}
 
-	svc.notify.NotifyBuildSuccess(ctx, app, deployment)
+	svc.notify.NotifyBuildSuccess(bgCtx, app, deployment)
 
 	// Deploy phase with timeout
-	err = svc.deployContainerWithTimeout(ctx, app, deployment, imageID)
+	err = svc.deployContainerWithTimeout(deployCtx, app, deployment, imageID)
 	if err != nil {
+		cancelErr := svc.checkCancelled(deployCtx, bgCtx, app, deployment)
+		if cancelErr != nil {
+			return cancelErr
+		}
+
 		return err
 	}
 
-	err = svc.updateAppRunning(ctx, app, imageID)
+	err = svc.updateAppRunning(bgCtx, app, imageID)
 	if err != nil {
 		return err
 	}
 
 	// Use context.WithoutCancel to ensure health check completes even if
 	// the parent context is cancelled (e.g., HTTP request ends).
-	go svc.checkHealthAfterDelay(context.WithoutCancel(ctx), app, deployment)
+	go svc.checkHealthAfterDelay(bgCtx, app, deployment)
 
 	return nil
 }
@@ -463,6 +534,43 @@ func (svc *Service) unlockApp(appID string) {
 	svc.getAppLock(appID).Unlock()
 }
 
+// cancelActiveDeploy cancels any in-progress deployment for the given app
+// and waits for it to finish before returning.
+func (svc *Service) cancelActiveDeploy(appID string) {
+	val, ok := svc.activeDeploys.Load(appID)
+	if !ok {
+		return
+	}
+
+	ad, ok := val.(*activeDeploy)
+	if !ok {
+		return
+	}
+
+	svc.log.Info("cancelling in-progress deployment", "app_id", appID)
+	ad.cancel()
+	<-ad.done
+}
+
+// checkCancelled checks if the deploy context was cancelled (by a newer deploy)
+// and if so, marks the deployment as cancelled. Returns ErrDeployCancelled or nil.
+func (svc *Service) checkCancelled(
+	deployCtx context.Context,
+	bgCtx context.Context,
+	app *models.App,
+	deployment *models.Deployment,
+) error {
+	if !errors.Is(deployCtx.Err(), context.Canceled) {
+		return nil
+	}
+
+	svc.log.Info("deployment cancelled by newer deploy", "app", app.Name)
+
+	_ = deployment.MarkFinished(bgCtx, models.DeploymentStatusCancelled)
+
+	return ErrDeployCancelled
+}
+
 func (svc *Service) fetchWebhookEvent(
 	ctx context.Context,
 	webhookEventID *int64,

internal/service/deploy/deploy_cancel_test.go

@@ -0,0 +1,133 @@
+package deploy_test
+
+import (
+	"context"
+	"log/slog"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+
+	"git.eeqj.de/sneak/upaas/internal/service/deploy"
+)
+
+func TestCancelActiveDeploy_NoExisting(t *testing.T) {
+	t.Parallel()
+
+	svc := deploy.NewTestService(slog.Default())
+
+	// Should not panic or block when no active deploy exists
+	svc.CancelActiveDeploy("nonexistent-app")
+}
+
+func TestCancelActiveDeploy_CancelsAndWaits(t *testing.T) {
+	t.Parallel()
+
+	svc := deploy.NewTestService(slog.Default())
+
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+
+	svc.RegisterActiveDeploy("app-1", cancel, done)
+
+	// Simulate a running deploy that respects cancellation
+	var deployFinished bool
+
+	go func() {
+		<-ctx.Done()
+
+		deployFinished = true
+
+		close(done)
+	}()
+
+	svc.CancelActiveDeploy("app-1")
+	assert.True(t, deployFinished, "deploy should have finished after cancellation")
+}
+
+func TestCancelActiveDeploy_BlocksUntilDone(t *testing.T) {
+	t.Parallel()
+
+	svc := deploy.NewTestService(slog.Default())
+
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+
+	svc.RegisterActiveDeploy("app-2", cancel, done)
+
+	// Simulate slow cleanup after cancellation
+	go func() {
+		<-ctx.Done()
+		time.Sleep(50 * time.Millisecond)
+		close(done)
+	}()
+
+	start := time.Now()
+
+	svc.CancelActiveDeploy("app-2")
+
+	elapsed := time.Since(start)
+
+	assert.GreaterOrEqual(t, elapsed, 50*time.Millisecond,
+		"cancelActiveDeploy should block until the deploy finishes")
+}
+
+func TestTryLockApp_PreventsConcurrent(t *testing.T) {
+	t.Parallel()
+
+	svc := deploy.NewTestService(slog.Default())
+
+	assert.True(t, svc.TryLockApp("app-1"), "first lock should succeed")
+	assert.False(t, svc.TryLockApp("app-1"), "second lock should fail")
+
+	svc.UnlockApp("app-1")
+
+	assert.True(t, svc.TryLockApp("app-1"), "lock after unlock should succeed")
+
+	svc.UnlockApp("app-1")
+}
+
+func TestCancelActiveDeploy_AllowsNewDeploy(t *testing.T) {
+	t.Parallel()
+
+	svc := deploy.NewTestService(slog.Default())
+
+	// Simulate an active deploy holding the lock
+	ctx, cancel := context.WithCancel(context.Background())
+	done := make(chan struct{})
+
+	svc.RegisterActiveDeploy("app-3", cancel, done)
+
+	// Lock the app as if a deploy is in progress
+	assert.True(t, svc.TryLockApp("app-3"))
+
+	// Simulate deploy goroutine: release lock on cancellation
+	var mu sync.Mutex
+
+	released := false
+
+	go func() {
+		<-ctx.Done()
+
+		svc.UnlockApp("app-3")
+
+		mu.Lock()
+		released = true
+		mu.Unlock()
+
+		close(done)
+	}()
+
+	// Cancel should cause the old deploy to release its lock
+	svc.CancelActiveDeploy("app-3")
+
+	mu.Lock()
+	assert.True(t, released)
+	mu.Unlock()
+
+	// Now a new deploy should be able to acquire the lock
+	assert.True(t, svc.TryLockApp("app-3"), "should be able to lock after cancellation")
+
+	svc.UnlockApp("app-3")
+}

internal/service/deploy/export_test.go

@@ -0,0 +1,33 @@
+package deploy
+
+import (
+	"context"
+	"log/slog"
+)
+
+// NewTestService creates a Service with minimal dependencies for testing.
+func NewTestService(log *slog.Logger) *Service {
+	return &Service{
+		log: log,
+	}
+}
+
+// CancelActiveDeploy exposes cancelActiveDeploy for testing.
+func (svc *Service) CancelActiveDeploy(appID string) {
+	svc.cancelActiveDeploy(appID)
+}
+
+// RegisterActiveDeploy registers an active deploy for testing.
+func (svc *Service) RegisterActiveDeploy(appID string, cancel context.CancelFunc, done chan struct{}) {
+	svc.activeDeploys.Store(appID, &activeDeploy{cancel: cancel, done: done})
+}
+
+// TryLockApp exposes tryLockApp for testing.
+func (svc *Service) TryLockApp(appID string) bool {
+	return svc.tryLockApp(appID)
+}
+
+// UnlockApp exposes unlockApp for testing.
+func (svc *Service) UnlockApp(appID string) {
+	svc.unlockApp(appID)
+}

internal/service/webhook/webhook.go

@@ -143,7 +143,7 @@ func (svc *Service) triggerDeployment(
 		// even if the HTTP request context is cancelled.
 		deployCtx := context.WithoutCancel(ctx)
 
-		deployErr := svc.deploy.Deploy(deployCtx, app, &eventID)
+		deployErr := svc.deploy.Deploy(deployCtx, app, &eventID, true)
 		if deployErr != nil {
 			svc.log.Error("deployment failed", "error", deployErr, "app", appName)
 		}

templates/app_detail.html

@@ -40,6 +40,10 @@
                     <span x-text="deploying ? 'Deploying...' : 'Deploy Now'"></span>
                 </button>
             </form>
+            <form method="POST" action="/apps/{{.App.ID}}/deployments/cancel" class="inline" x-show="deploying" x-cloak x-data="confirmAction('Cancel the current deployment?')" @submit="confirm($event)">
+                {{ .CSRFField }}
+                <button type="submit" class="btn-danger">Cancel Deploy</button>
+            </form>
         </div>
     </div>