feat: add resilient container restart controls

internal/web/server.go

@@ -33,16 +33,20 @@ import (
 var embeddedAssets embed.FS
 
 type Server struct {
-	config    config.Config
-	store     *store.Store
-	cipher    *secure.Cipher
-	scheduler *scheduler.Scheduler
-	updates   *appupdate.Manager
-	logger    *slog.Logger
-	templates map[string]*template.Template
-	limiter   *loginLimiter
-	dummyHash string
-	authSlots chan struct{}
+	config           config.Config
+	store            *store.Store
+	cipher           *secure.Cipher
+	scheduler        *scheduler.Scheduler
+	updates          *appupdate.Manager
+	logger           *slog.Logger
+	templates        map[string]*template.Template
+	limiter          *loginLimiter
+	dummyHash        string
+	authSlots        chan struct{}
+	instanceID       string
+	restartContainer func()
+	restartMu        sync.Mutex
+	restartPending   bool
 }
 
 type viewData struct {
@@ -79,6 +83,7 @@ type viewData struct {
 	ErrorCount       int
 	PiHoleView       *piHoleView
 	SoftwareUpdate   appupdate.State
+	InstanceID       string
 }
 
 type flashMessage struct {
@@ -113,12 +118,16 @@ type schedulerResponse struct {
 	RecentRuns  *[]schedulerRunResponse `json:"recent_runs,omitempty"`
 }
 
-func New(cfg config.Config, database *store.Store, cipher *secure.Cipher, schedule *scheduler.Scheduler, updates *appupdate.Manager, logger *slog.Logger) (*Server, error) {
+func New(cfg config.Config, database *store.Store, cipher *secure.Cipher, schedule *scheduler.Scheduler, updates *appupdate.Manager, logger *slog.Logger, restartContainer func()) (*Server, error) {
 	templates, err := loadTemplates()
 	if err != nil {
 		return nil, err
 	}
-	return &Server{config: cfg, store: database, cipher: cipher, scheduler: schedule, updates: updates, logger: logger, templates: templates, limiter: newLoginLimiter(), dummyHash: secure.PasswordTimingHash, authSlots: make(chan struct{}, 2)}, nil
+	instanceID, err := secure.RandomToken(18)
+	if err != nil {
+		return nil, fmt.Errorf("create process instance ID: %w", err)
+	}
+	return &Server{config: cfg, store: database, cipher: cipher, scheduler: schedule, updates: updates, logger: logger, templates: templates, limiter: newLoginLimiter(), dummyHash: secure.PasswordTimingHash, authSlots: make(chan struct{}, 2), instanceID: instanceID, restartContainer: restartContainer}, nil
 }
 
 func (s *Server) Handler() http.Handler {
@@ -170,6 +179,7 @@ func (s *Server) Handler() http.Handler {
 	mux.Handle("POST /system/update-credential", s.adminOnly(s.softwareUpdateCredentialSave))
 	mux.Handle("POST /system/check", s.adminOnly(s.softwareUpdateCheck))
 	mux.Handle("POST /system/update", s.adminOnly(s.softwareUpdateStart))
+	mux.Handle("POST /system/restart", s.adminOnly(s.containerRestart))
 	mux.Handle("GET /errors", s.adminOnly(s.errorsPage))
 	mux.Handle("POST /errors/clear", s.adminOnly(s.errorsClear))
 	mux.Handle("GET /account", s.authenticated(s.accountPage))
@@ -192,6 +202,10 @@ func (s *Server) Handler() http.Handler {
 
 func (s *Server) health(w http.ResponseWriter, _ *http.Request) {
 	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
+	w.Header().Set("Cache-Control", "no-store")
+	if s.instanceID != "" {
+		w.Header().Set("X-Domain-Control-Instance", s.instanceID)
+	}
 	w.WriteHeader(http.StatusOK)
 	_, _ = w.Write([]byte("ok\n"))
 }
@@ -1304,7 +1318,7 @@ func (s *Server) softwareUpdatePage(w http.ResponseWriter, r *http.Request, sess
 	statusCtx, cancel := context.WithTimeout(r.Context(), 750*time.Millisecond)
 	_, _ = s.updates.UpdateStatus(statusCtx)
 	cancel()
-	s.render(w, "system", viewData{Title: "Systemupdate", Page: "system", CSRF: session.CSRFToken, Flash: notice(r), User: userFromSession(session), IsAdmin: true, Now: time.Now(), SoftwareUpdate: s.updates.State()})
+	s.render(w, "system", viewData{Title: "Systemupdate", Page: "system", CSRF: session.CSRFToken, Flash: notice(r), User: userFromSession(session), IsAdmin: true, Now: time.Now(), SoftwareUpdate: s.updates.State(), InstanceID: s.instanceID})
 }
 
 func (s *Server) softwareUpdateCredentialSave(w http.ResponseWriter, r *http.Request, session *model.Session) {
@@ -1418,6 +1432,63 @@ func (s *Server) softwareUpdateStatusAPI(w http.ResponseWriter, r *http.Request,
 	})
 }
 
+func (s *Server) containerRestart(w http.ResponseWriter, r *http.Request, session *model.Session) {
+	if err := s.parseForm(w, r); err != nil {
+		return
+	}
+	if !s.validSessionCSRF(r, session) {
+		http.Error(w, "Ungültige Anfrage", http.StatusForbidden)
+		return
+	}
+	if s.restartContainer == nil {
+		http.Error(w, "Container-Neustart ist nicht konfiguriert.", http.StatusServiceUnavailable)
+		return
+	}
+	if s.updates != nil {
+		statusCtx, cancel := context.WithTimeout(r.Context(), time.Second)
+		updateStatus, statusErr := s.updates.UpdateStatus(statusCtx)
+		cancel()
+		if statusErr == nil && updateStatus.Busy() {
+			http.Error(w, "Während eines Software-Updates kann der Container nicht zusätzlich neu gestartet werden.", http.StatusConflict)
+			return
+		}
+	}
+
+	s.restartMu.Lock()
+	if s.restartPending {
+		s.restartMu.Unlock()
+		s.writeRestartAccepted(w, r, true)
+		return
+	}
+	if err := s.store.AddAudit(r.Context(), "system.container_restart", "Container-Neustart angefordert durch "+session.Username); err != nil {
+		s.restartMu.Unlock()
+		s.serverError(w, err)
+		return
+	}
+	s.restartPending = true
+	restart := s.restartContainer
+	s.restartMu.Unlock()
+
+	s.writeRestartAccepted(w, r, false)
+	if flusher, ok := w.(http.Flusher); ok {
+		flusher.Flush()
+	}
+	restart()
+}
+
+func (s *Server) writeRestartAccepted(w http.ResponseWriter, r *http.Request, alreadyPending bool) {
+	w.Header().Set("Cache-Control", "no-store")
+	if strings.Contains(r.Header.Get("Accept"), "application/json") {
+		w.Header().Set("Content-Type", "application/json; charset=utf-8")
+		w.WriteHeader(http.StatusAccepted)
+		_ = json.NewEncoder(w).Encode(map[string]any{"accepted": true, "already_pending": alreadyPending, "instance_id": s.instanceID})
+		return
+	}
+	w.Header().Set("Content-Type", "text/plain; charset=utf-8")
+	w.WriteHeader(http.StatusAccepted)
+	_, _ = w.Write([]byte("Container-Neustart angefordert. Diese Seite ist kurz nicht erreichbar; lade sie anschließend neu.\n"))
+}
+
 func (s *Server) errorsPage(w http.ResponseWriter, r *http.Request, session *model.Session) {
 	targets, err := s.store.ListTargetsWithErrors(r.Context())
 	if err != nil {