Merge pull request #78 from lordmathis/feat/inflight-requests

feat: Wait for inflight requests to finish before shutting down an instance

cmd/server/main.go

@@ -5,6 +5,7 @@ import (
 	"llamactl/pkg/config"
 	"llamactl/pkg/manager"
 	"llamactl/pkg/server"
+	"log"
 	"net/http"
 	"os"
 	"os/signal"
@@ -38,8 +39,7 @@ func main() {
 	configPath := os.Getenv("LLAMACTL_CONFIG_PATH")
 	cfg, err := config.LoadConfig(configPath)
 	if err != nil {
-		fmt.Printf("Error loading config: %v\n", err)
+		log.Printf("Error loading config: %v\nUsing default configuration.", err)
-		fmt.Println("Using default configuration.")
 	}
 
 	// Set version information
@@ -50,13 +50,11 @@ func main() {
 	// Create the data directory if it doesn't exist
 	if cfg.Instances.AutoCreateDirs {
 		if err := os.MkdirAll(cfg.Instances.InstancesDir, 0755); err != nil {
-			fmt.Printf("Error creating config directory %s: %v\n", cfg.Instances.InstancesDir, err)
+			log.Printf("Error creating config directory %s: %v\nPersistence will not be available.", cfg.Instances.InstancesDir, err)
-			fmt.Println("Persistence will not be available.")
 		}
 
 		if err := os.MkdirAll(cfg.Instances.LogsDir, 0755); err != nil {
-			fmt.Printf("Error creating log directory %s: %v\n", cfg.Instances.LogsDir, err)
+			log.Printf("Error creating log directory %s: %v\nInstance logs will not be available.", cfg.Instances.LogsDir, err)
-			fmt.Println("Instance logs will not be available.")
 		}
 	}
 
@@ -81,7 +79,7 @@ func main() {
 	go func() {
 		fmt.Printf("Llamactl server listening on %s:%d\n", cfg.Server.Host, cfg.Server.Port)
 		if err := server.ListenAndServe(); err != nil && err != http.ErrServerClosed {
-			fmt.Printf("Error starting server: %v\n", err)
+			log.Printf("Error starting server: %v\n", err)
 		}
 	}()
 
@@ -90,7 +88,7 @@ func main() {
 	fmt.Println("Shutting down server...")
 
 	if err := server.Close(); err != nil {
-		fmt.Printf("Error shutting down server: %v\n", err)
+		log.Printf("Error shutting down server: %v\n", err)
 	} else {
 		fmt.Println("Server shut down gracefully.")
 	}

docs/managing-instances.md

@@ -42,6 +42,7 @@ Each instance is displayed as a card showing:
 ![Create Instance Screenshot](images/create_instance.png)
 
 1. Click the **"Create Instance"** button on the dashboard
+2. *Optional*: Click **"Import"** in the dialog header to load a previously exported configuration
 2. Enter a unique **Name** for your instance (only required field)
 3. **Select Target Node**: Choose which node to deploy the instance to from the dropdown
 4. **Choose Backend Type**:
@@ -219,6 +220,12 @@ curl -X PUT http://localhost:8080/api/v1/instances/{name} \
     Configuration changes require restarting the instance to take effect.
 
 
+## Export Instance
+
+**Via Web UI**
+1. Click the **"More actions"** button (three dots) on an instance card
+2. Click **"Export"** to download the instance configuration as a JSON file
+
 ## View Logs
 
 **Via Web UI**

pkg/backends/backend.go

@@ -93,6 +93,8 @@ func (o *Options) MarshalJSON() ([]byte, error) {
 		if err != nil {
 			return nil, fmt.Errorf("failed to marshal backend options: %w", err)
 		}
+		// Create a new map to avoid concurrent map writes
+		aux.BackendOptions = make(map[string]any)
 		if err := json.Unmarshal(optionsData, &aux.BackendOptions); err != nil {
 			return nil, fmt.Errorf("failed to unmarshal backend options to map: %w", err)
 		}

pkg/config/config.go

@@ -1,6 +1,7 @@
 package config
 
 import (
+	"fmt"
 	"log"
 	"os"
 	"path/filepath"
@@ -231,6 +232,11 @@ func LoadConfig(configPath string) (AppConfig, error) {
 		cfg.Instances.LogsDir = filepath.Join(cfg.Instances.DataDir, "logs")
 	}
 
+	// Validate port range
+	if cfg.Instances.PortRange[0] <= 0 || cfg.Instances.PortRange[1] <= 0 || cfg.Instances.PortRange[0] >= cfg.Instances.PortRange[1] {
+		return AppConfig{}, fmt.Errorf("invalid port range: %v", cfg.Instances.PortRange)
+	}
+
 	return cfg, nil
 }
 

pkg/instance/instance.go

@@ -5,7 +5,7 @@ import (
 	"fmt"
 	"llamactl/pkg/config"
 	"log"
-	"net/http/httputil"
+	"net/http"
 	"time"
 )
 
@@ -182,15 +182,6 @@ func (i *Instance) GetPort() int {
 	return i.options.GetPort()
 }
 
-// GetProxy returns the reverse proxy for this instance
-func (i *Instance) GetProxy() (*httputil.ReverseProxy, error) {
-	if i.proxy == nil {
-		return nil, fmt.Errorf("instance %s has no proxy component", i.Name)
-	}
-
-	return i.proxy.get()
-}
-
 func (i *Instance) IsRemote() bool {
 	opts := i.GetOptions()
 	if opts == nil {
@@ -242,6 +233,22 @@ func (i *Instance) ShouldTimeout() bool {
 	return i.proxy.shouldTimeout()
 }
 
+// GetInflightRequests returns the current number of inflight requests
+func (i *Instance) GetInflightRequests() int32 {
+	if i.proxy == nil {
+		return 0
+	}
+	return i.proxy.getInflightRequests()
+}
+
+// ServeHTTP serves HTTP requests through the proxy with request tracking and shutdown handling
+func (i *Instance) ServeHTTP(w http.ResponseWriter, r *http.Request) error {
+	if i.proxy == nil {
+		return fmt.Errorf("instance %s has no proxy component", i.Name)
+	}
+	return i.proxy.serveHTTP(w, r)
+}
+
 func (i *Instance) getCommand() string {
 	opts := i.GetOptions()
 	if opts == nil {

pkg/instance/instance_test.go

@@ -171,64 +171,6 @@ func TestSetOptions(t *testing.T) {
 	}
 }
 
-func TestGetProxy(t *testing.T) {
-	globalConfig := &config.AppConfig{
-		Backends: config.BackendConfig{
-			LlamaCpp: config.BackendSettings{
-				Command: "llama-server",
-				Args:    []string{},
-			},
-			MLX: config.BackendSettings{
-				Command: "mlx_lm.server",
-				Args:    []string{},
-			},
-			VLLM: config.BackendSettings{
-				Command: "vllm",
-				Args:    []string{"serve"},
-			},
-		},
-		Instances: config.InstancesConfig{
-			LogsDir: "/tmp/test",
-		},
-		Nodes:     map[string]config.NodeConfig{},
-		LocalNode: "main",
-	}
-
-	options := &instance.Options{
-		Nodes: map[string]struct{}{"main": {}},
-		BackendOptions: backends.Options{
-			BackendType: backends.BackendTypeLlamaCpp,
-			LlamaServerOptions: &backends.LlamaServerOptions{
-				Host: "localhost",
-				Port: 8080,
-			},
-		},
-	}
-
-	// Mock onStatusChange function
-	mockOnStatusChange := func(oldStatus, newStatus instance.Status) {}
-
-	inst := instance.New("test-instance", globalConfig, options, mockOnStatusChange)
-
-	// Get proxy for the first time
-	proxy1, err := inst.GetProxy()
-	if err != nil {
-		t.Fatalf("GetProxy failed: %v", err)
-	}
-	if proxy1 == nil {
-		t.Error("Expected proxy to be created")
-	}
-
-	// Get proxy again - should return cached version
-	proxy2, err := inst.GetProxy()
-	if err != nil {
-		t.Fatalf("GetProxy failed: %v", err)
-	}
-	if proxy1 != proxy2 {
-		t.Error("Expected cached proxy to be returned")
-	}
-}
-
 func TestMarshalJSON(t *testing.T) {
 	globalConfig := &config.AppConfig{
 		Backends: config.BackendConfig{
@@ -613,11 +555,6 @@ func TestRemoteInstanceOperations(t *testing.T) {
 		t.Error("Expected error when restarting remote instance")
 	}
 
-	// GetProxy should fail for remote instance
-	if _, err := inst.GetProxy(); err != nil {
-		t.Error("Expected no error when getting proxy for remote instance")
-	}
-
 	// GetLogs should fail for remote instance
 	if _, err := inst.GetLogs(10); err == nil {
 		t.Error("Expected error when getting logs for remote instance")

pkg/instance/logger.go

@@ -7,13 +7,14 @@ import (
 	"os"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 )
 
 type logger struct {
 	name        string
 	logDir      string
-	logFile     *os.File
+	logFile     atomic.Pointer[os.File]
 	logFilePath string
 	mu          sync.RWMutex
 }
@@ -47,11 +48,11 @@ func (i *logger) create() error {
 		return fmt.Errorf("failed to create stdout log file: %w", err)
 	}
 
-	i.logFile = logFile
+	i.logFile.Store(logFile)
 
 	// Write a startup marker to both files
 	timestamp := time.Now().Format("2006-01-02 15:04:05")
-	fmt.Fprintf(i.logFile, "\n=== Instance %s started at %s ===\n", i.name, timestamp)
+	fmt.Fprintf(logFile, "\n=== Instance %s started at %s ===\n", i.name, timestamp)
 
 	return nil
 }
@@ -102,11 +103,12 @@ func (i *logger) close() {
 	i.mu.Lock()
 	defer i.mu.Unlock()
 
-	if i.logFile != nil {
+	logFile := i.logFile.Swap(nil)
+	if logFile != nil {
 		timestamp := time.Now().Format("2006-01-02 15:04:05")
-		fmt.Fprintf(i.logFile, "=== Instance %s stopped at %s ===\n\n", i.name, timestamp)
+		fmt.Fprintf(logFile, "=== Instance %s stopped at %s ===\n\n", i.name, timestamp)
-		i.logFile.Close()
+		logFile.Sync() // Ensure all buffered data is written to disk
-		i.logFile = nil
+		logFile.Close()
 	}
 }
 
@@ -117,9 +119,9 @@ func (i *logger) readOutput(reader io.ReadCloser) {
 	scanner := bufio.NewScanner(reader)
 	for scanner.Scan() {
 		line := scanner.Text()
-		if i.logFile != nil {
+		// Use atomic load to avoid lock contention on every line
-			fmt.Fprintln(i.logFile, line)
+		if logFile := i.logFile.Load(); logFile != nil {
-			i.logFile.Sync() // Ensure data is written to disk
+			fmt.Fprintln(logFile, line)
 		}
 	}
 }

pkg/instance/process.go

@@ -132,14 +132,28 @@ func (p *process) stop() error {
 		p.restartCancel = nil
 	}
 
-	// Set status to stopped first to signal intentional stop
+	// Set status to ShuttingDown first to reject new requests
-	p.instance.SetStatus(Stopped)
+	p.instance.SetStatus(ShuttingDown)
 
 	// Get the monitor done channel before releasing the lock
 	monitorDone := p.monitorDone
 
 	p.mu.Unlock()
 
+	// Wait for inflight requests to complete (max 30 seconds)
+	log.Printf("Instance %s shutting down, waiting for inflight requests to complete...", p.instance.Name)
+	deadline := time.Now().Add(30 * time.Second)
+	for time.Now().Before(deadline) {
+		inflight := p.instance.GetInflightRequests()
+		if inflight == 0 {
+			break
+		}
+		time.Sleep(100 * time.Millisecond)
+	}
+
+	// Now set status to stopped to signal intentional stop
+	p.instance.SetStatus(Stopped)
+
 	// Stop the process with SIGINT if cmd exists
 	if p.cmd != nil && p.cmd.Process != nil {
 		if err := p.cmd.Process.Signal(syscall.SIGINT); err != nil {
@@ -156,6 +170,7 @@ func (p *process) stop() error {
 	select {
 	case <-monitorDone:
 		// Process exited normally
+		log.Printf("Instance %s shut down gracefully", p.instance.Name)
 	case <-time.After(30 * time.Second):
 		// Force kill if it doesn't exit within 30 seconds
 		if p.cmd != nil && p.cmd.Process != nil {

pkg/instance/proxy.go

@@ -37,8 +37,9 @@ type proxy struct {
 	proxyOnce sync.Once
 	proxyErr  error
 
-	lastRequestTime atomic.Int64
+	lastRequestTime  atomic.Int64
-	timeProvider    TimeProvider
+	inflightRequests atomic.Int32
+	timeProvider     TimeProvider
 }
 
 // newProxy creates a new Proxy for the given instance
@@ -153,6 +154,23 @@ func (p *proxy) build() (*httputil.ReverseProxy, error) {
 	return proxy, nil
 }
 
+// serveHTTP handles HTTP requests with inflight tracking
+func (p *proxy) serveHTTP(w http.ResponseWriter, r *http.Request) error {
+	// Get the reverse proxy
+	reverseProxy, err := p.get()
+	if err != nil {
+		return err
+	}
+
+	// Track inflight requests
+	p.incInflightRequests()
+	defer p.decInflightRequests()
+
+	// Serve the request
+	reverseProxy.ServeHTTP(w, r)
+	return nil
+}
+
 // clear resets the proxy, allowing it to be recreated when options change.
 func (p *proxy) clear() {
 	p.mu.Lock()
@@ -160,7 +178,7 @@ func (p *proxy) clear() {
 
 	p.proxy = nil
 	p.proxyErr = nil
-	p.proxyOnce = sync.Once{} // Reset Once for next GetProxy call
+	p.proxyOnce = sync.Once{}
 }
 
 // updateLastRequestTime updates the last request access time for the instance
@@ -199,3 +217,18 @@ func (p *proxy) shouldTimeout() bool {
 func (p *proxy) setTimeProvider(tp TimeProvider) {
 	p.timeProvider = tp
 }
+
+// incInflightRequests increments the inflight request counter
+func (p *proxy) incInflightRequests() {
+	p.inflightRequests.Add(1)
+}
+
+// decInflightRequests decrements the inflight request counter
+func (p *proxy) decInflightRequests() {
+	p.inflightRequests.Add(-1)
+}
+
+// getInflightRequests returns the current number of inflight requests
+func (p *proxy) getInflightRequests() int32 {
+	return p.inflightRequests.Load()
+}

pkg/instance/status.go

@@ -14,20 +14,23 @@ const (
 	Running
 	Failed
 	Restarting
+	ShuttingDown
 )
 
 var nameToStatus = map[string]Status{
-	"stopped":    Stopped,
+	"stopped":       Stopped,
-	"running":    Running,
+	"running":       Running,
-	"failed":     Failed,
+	"failed":        Failed,
-	"restarting": Restarting,
+	"restarting":    Restarting,
+	"shutting_down": ShuttingDown,
 }
 
 var statusToName = map[Status]string{
-	Stopped:    "stopped",
+	Stopped:      "stopped",
-	Running:    "running",
+	Running:      "running",
-	Failed:     "failed",
+	Failed:       "failed",
-	Restarting: "restarting",
+	Restarting:   "restarting",
+	ShuttingDown: "shutting_down",
 }
 
 // Status enum JSON marshaling methods

pkg/manager/lifecycle_test.go

@@ -10,7 +10,7 @@ import (
 )
 
 func TestInstanceTimeoutLogic(t *testing.T) {
-	testManager := createTestManager()
+	testManager := createTestManager(t)
 	defer testManager.Shutdown()
 
 	idleTimeout := 1 // 1 minute
@@ -42,7 +42,7 @@ func TestInstanceTimeoutLogic(t *testing.T) {
 }
 
 func TestInstanceWithoutTimeoutNeverExpires(t *testing.T) {
-	testManager := createTestManager()
+	testManager := createTestManager(t)
 	defer testManager.Shutdown()
 
 	noTimeoutInst := createInstanceWithTimeout(t, testManager, "no-timeout-test", "/path/to/model.gguf", nil)
@@ -64,7 +64,7 @@ func TestInstanceWithoutTimeoutNeverExpires(t *testing.T) {
 }
 
 func TestEvictLRUInstance_Success(t *testing.T) {
-	manager := createTestManager()
+	manager := createTestManager(t)
 	defer manager.Shutdown()
 
 	// Create 3 instances with idle timeout enabled (value doesn't matter for LRU logic)
@@ -121,7 +121,7 @@ func TestEvictLRUInstance_Success(t *testing.T) {
 }
 
 func TestEvictLRUInstance_NoRunningInstances(t *testing.T) {
-	manager := createTestManager()
+	manager := createTestManager(t)
 	defer manager.Shutdown()
 
 	err := manager.EvictLRUInstance()
@@ -134,7 +134,7 @@ func TestEvictLRUInstance_NoRunningInstances(t *testing.T) {
 }
 
 func TestEvictLRUInstance_OnlyEvictsTimeoutEnabledInstances(t *testing.T) {
-	manager := createTestManager()
+	manager := createTestManager(t)
 	defer manager.Shutdown()
 
 	// Create mix of instances: some with timeout enabled, some disabled

pkg/manager/manager.go

@@ -54,16 +54,10 @@ func New(globalConfig *config.AppConfig) InstanceManager {
 
 	// Initialize port allocator
 	portRange := globalConfig.Instances.PortRange
-	ports, err := newPortAllocator(portRange[0], portRange[1])
+	ports := newPortAllocator(portRange[0], portRange[1])
-	if err != nil {
-		log.Fatalf("Failed to create port allocator: %v", err)
-	}
 
 	// Initialize persistence
-	persistence, err := newInstancePersister(globalConfig.Instances.InstancesDir)
+	persistence := newInstancePersister(globalConfig.Instances.InstancesDir)
-	if err != nil {
-		log.Fatalf("Failed to create instance persister: %v", err)
-	}
 
 	// Initialize remote manager
 	remote := newRemoteManager(globalConfig.Nodes, 30*time.Second)
@@ -116,7 +110,7 @@ func (im *instanceManager) Shutdown() {
 				defer wg.Done()
 				fmt.Printf("Stopping instance %s...\n", inst.Name)
 				if err := inst.Stop(); err != nil {
-					fmt.Printf("Error stopping instance %s: %v\n", inst.Name, err)
+					log.Printf("Error stopping instance %s: %v\n", inst.Name, err)
 				}
 			}(inst)
 		}

pkg/manager/manager_test.go

@@ -85,7 +85,7 @@ func TestDeleteInstance_RemovesPersistenceFile(t *testing.T) {
 }
 
 func TestConcurrentAccess(t *testing.T) {
-	mgr := createTestManager()
+	mgr := createTestManager(t)
 	defer mgr.Shutdown()
 
 	// Test concurrent operations
@@ -113,7 +113,7 @@ func TestConcurrentAccess(t *testing.T) {
 	}
 
 	// Concurrent list operations
-	for i := 0; i < 3; i++ {
+	for range 3 {
 		wg.Add(1)
 		go func() {
 			defer wg.Done()
@@ -134,16 +134,17 @@ func TestConcurrentAccess(t *testing.T) {
 
 // Helper functions for test configuration
 func createTestAppConfig(instancesDir string) *config.AppConfig {
-	// Use 'sleep' as a test command instead of 'llama-server'
+	// Use 'sh -c "sleep 999999"' as a test command instead of 'llama-server'
-	// This allows tests to run in CI environments without requiring actual LLM binaries
+	// The shell ignores all additional arguments passed after the command
-	// The sleep command will be invoked with model paths and other args, which it ignores
 	return &config.AppConfig{
 		Backends: config.BackendConfig{
 			LlamaCpp: config.BackendSettings{
-				Command: "sleep",
+				Command: "sh",
+				Args:    []string{"-c", "sleep 999999"},
 			},
 			MLX: config.BackendSettings{
-				Command: "sleep",
+				Command: "sh",
+				Args:    []string{"-c", "sleep 999999"},
 			},
 		},
 		Instances: config.InstancesConfig{
@@ -162,28 +163,8 @@ func createTestAppConfig(instancesDir string) *config.AppConfig {
 	}
 }
 
-func createTestManager() manager.InstanceManager {
+func createTestManager(t *testing.T) manager.InstanceManager {
-	appConfig := &config.AppConfig{
+	tempDir := t.TempDir()
-		Backends: config.BackendConfig{
+	appConfig := createTestAppConfig(tempDir)
-			LlamaCpp: config.BackendSettings{
-				Command: "sleep",
-			},
-			MLX: config.BackendSettings{
-				Command: "sleep",
-			},
-		},
-		Instances: config.InstancesConfig{
-			PortRange:            [2]int{8000, 9000},
-			LogsDir:              "/tmp/test",
-			MaxInstances:         10,
-			MaxRunningInstances:  10,
-			DefaultAutoRestart:   true,
-			DefaultMaxRestarts:   3,
-			DefaultRestartDelay:  5,
-			TimeoutCheckInterval: 5,
-		},
-		LocalNode: "main",
-		Nodes:     map[string]config.NodeConfig{},
-	}
 	return manager.New(appConfig)
 }

pkg/manager/operations.go

@@ -330,7 +330,8 @@ func (im *instanceManager) DeleteInstance(name string) error {
 	lock.Lock()
 	defer im.unlockAndCleanup(name)
 
-	if inst.IsRunning() {
+	status := inst.GetStatus()
+	if status == instance.Running || status == instance.Restarting {
 		return fmt.Errorf("instance with name %s is still running, stop it before deleting", name)
 	}
 

pkg/manager/operations_test.go

@@ -10,7 +10,7 @@ import (
 )
 
 func TestCreateInstance_FailsWithDuplicateName(t *testing.T) {
-	mngr := createTestManager()
+	mngr := createTestManager(t)
 	options := &instance.Options{
 		BackendOptions: backends.Options{
 			BackendType: backends.BackendTypeLlamaCpp,
@@ -36,6 +36,7 @@ func TestCreateInstance_FailsWithDuplicateName(t *testing.T) {
 }
 
 func TestCreateInstance_FailsWhenMaxInstancesReached(t *testing.T) {
+	tempDir := t.TempDir()
 	appConfig := &config.AppConfig{
 		Backends: config.BackendConfig{
 			LlamaCpp: config.BackendSettings{
@@ -44,6 +45,7 @@ func TestCreateInstance_FailsWhenMaxInstancesReached(t *testing.T) {
 		},
 		Instances: config.InstancesConfig{
 			PortRange:            [2]int{8000, 9000},
+			InstancesDir:         tempDir,
 			MaxInstances:         1, // Very low limit for testing
 			TimeoutCheckInterval: 5,
 		},
@@ -77,7 +79,7 @@ func TestCreateInstance_FailsWhenMaxInstancesReached(t *testing.T) {
 }
 
 func TestCreateInstance_FailsWithPortConflict(t *testing.T) {
-	manager := createTestManager()
+	manager := createTestManager(t)
 
 	options1 := &instance.Options{
 		BackendOptions: backends.Options{
@@ -115,7 +117,7 @@ func TestCreateInstance_FailsWithPortConflict(t *testing.T) {
 }
 
 func TestInstanceOperations_FailWithNonExistentInstance(t *testing.T) {
-	manager := createTestManager()
+	manager := createTestManager(t)
 
 	options := &instance.Options{
 		BackendOptions: backends.Options{
@@ -143,7 +145,7 @@ func TestInstanceOperations_FailWithNonExistentInstance(t *testing.T) {
 }
 
 func TestDeleteInstance_RunningInstanceFails(t *testing.T) {
-	mgr := createTestManager()
+	mgr := createTestManager(t)
 	defer mgr.Shutdown()
 
 	options := &instance.Options{
@@ -155,15 +157,13 @@ func TestDeleteInstance_RunningInstanceFails(t *testing.T) {
 		},
 	}
 
-	_, err := mgr.CreateInstance("test-instance", options)
+	inst, err := mgr.CreateInstance("test-instance", options)
 	if err != nil {
 		t.Fatalf("CreateInstance failed: %v", err)
 	}
 
-	_, err = mgr.StartInstance("test-instance")
+	// Simulate starting the instance
-	if err != nil {
+	inst.SetStatus(instance.Running)
-		t.Fatalf("StartInstance failed: %v", err)
-	}
 
 	// Should fail to delete running instance
 	err = mgr.DeleteInstance("test-instance")
@@ -173,7 +173,7 @@ func TestDeleteInstance_RunningInstanceFails(t *testing.T) {
 }
 
 func TestUpdateInstance(t *testing.T) {
-	mgr := createTestManager()
+	mgr := createTestManager(t)
 	defer mgr.Shutdown()
 
 	options := &instance.Options{
@@ -186,14 +186,14 @@ func TestUpdateInstance(t *testing.T) {
 		},
 	}
 
-	_, err := mgr.CreateInstance("test-instance", options)
+	inst, err := mgr.CreateInstance("test-instance", options)
 	if err != nil {
 		t.Fatalf("CreateInstance failed: %v", err)
 	}
 
-	_, err = mgr.StartInstance("test-instance")
+	// Start the instance (will use 'yes' command from test config)
-	if err != nil {
+	if err := inst.Start(); err != nil {
-		t.Fatalf("StartInstance failed: %v", err)
+		t.Fatalf("Failed to start instance: %v", err)
 	}
 
 	// Update running instance with new model
@@ -212,9 +212,9 @@ func TestUpdateInstance(t *testing.T) {
 		t.Fatalf("UpdateInstance failed: %v", err)
 	}
 
-	// Should still be running after update
+	// Should be running after update (was running before, should be restarted)
 	if !updated.IsRunning() {
-		t.Error("Instance should be running after update")
+		t.Errorf("Instance should be running after update, got: %v", updated.GetStatus())
 	}
 
 	if updated.GetOptions().BackendOptions.LlamaServerOptions.Model != "/path/to/new-model.gguf" {
@@ -223,7 +223,7 @@ func TestUpdateInstance(t *testing.T) {
 }
 
 func TestUpdateInstance_ReleasesOldPort(t *testing.T) {
-	mgr := createTestManager()
+	mgr := createTestManager(t)
 	defer mgr.Shutdown()
 
 	options := &instance.Options{

pkg/manager/persistence.go

@@ -15,35 +15,18 @@ import (
 type instancePersister struct {
 	mu           sync.Mutex
 	instancesDir string
-	enabled      bool
 }
 
 // newInstancePersister creates a new instance persister.
 // If instancesDir is empty, persistence is disabled.
-func newInstancePersister(instancesDir string) (*instancePersister, error) {
+func newInstancePersister(instancesDir string) *instancePersister {
-	if instancesDir == "" {
-		return &instancePersister{
-			enabled: false,
-		}, nil
-	}
-
-	// Ensure the instances directory exists
-	if err := os.MkdirAll(instancesDir, 0755); err != nil {
-		return nil, fmt.Errorf("failed to create instances directory: %w", err)
-	}
-
 	return &instancePersister{
 		instancesDir: instancesDir,
-		enabled:      true,
+	}
-	}, nil
 }
 
 // Save persists an instance to disk with atomic write
 func (p *instancePersister) save(inst *instance.Instance) error {
-	if !p.enabled {
-		return nil
-	}
-
 	if inst == nil {
 		return fmt.Errorf("cannot save nil instance")
 	}
@@ -103,10 +86,6 @@ func (p *instancePersister) save(inst *instance.Instance) error {
 
 // Delete removes an instance's persistence file from disk.
 func (p *instancePersister) delete(name string) error {
-	if !p.enabled {
-		return nil
-	}
-
 	validatedName, err := p.validateInstanceName(name)
 	if err != nil {
 		return err
@@ -131,10 +110,6 @@ func (p *instancePersister) delete(name string) error {
 // LoadAll loads all persisted instances from disk.
 // Returns a slice of instances and any errors encountered during loading.
 func (p *instancePersister) loadAll() ([]*instance.Instance, error) {
-	if !p.enabled {
-		return nil, nil
-	}
-
 	p.mu.Lock()
 	defer p.mu.Unlock()
 

pkg/manager/ports.go

@@ -24,15 +24,7 @@ type portAllocator struct {
 }
 
 // newPortAllocator creates a new port allocator for the given port range.
-// Returns an error if the port range is invalid.
+func newPortAllocator(minPort, maxPort int) *portAllocator {
-func newPortAllocator(minPort, maxPort int) (*portAllocator, error) {
-	if minPort <= 0 || maxPort <= 0 {
-		return nil, fmt.Errorf("invalid port range: min=%d, max=%d (must be > 0)", minPort, maxPort)
-	}
-	if minPort > maxPort {
-		return nil, fmt.Errorf("invalid port range: min=%d > max=%d", minPort, maxPort)
-	}
-
 	rangeSize := maxPort - minPort + 1
 	bitmapSize := (rangeSize + 63) / 64 // Round up to nearest uint64
 
@@ -42,7 +34,7 @@ func newPortAllocator(minPort, maxPort int) (*portAllocator, error) {
 		minPort:   minPort,
 		maxPort:   maxPort,
 		rangeSize: rangeSize,
-	}, nil
+	}
 }
 
 // allocate finds and allocates the first available port for the given instance.

pkg/server/handlers_backends.go

@@ -66,17 +66,16 @@ func (h *Handler) LlamaCppUIProxy() http.HandlerFunc {
 			return
 		}
 
-		proxy, err := inst.GetProxy()
-		if err != nil {
-			writeError(w, http.StatusInternalServerError, "failed to get proxy", err.Error())
-			return
-		}
-
 		if !inst.IsRemote() {
 			h.stripLlamaCppPrefix(r, inst.Name)
 		}
 
-		proxy.ServeHTTP(w, r)
+		// Use instance's ServeHTTP which tracks inflight requests and handles shutting down state
+		err = inst.ServeHTTP(w, r)
+		if err != nil {
+			// Error is already handled in ServeHTTP (response written)
+			return
+		}
 	}
 }
 
@@ -110,6 +109,12 @@ func (h *Handler) LlamaCppProxy() http.HandlerFunc {
 			return
 		}
 
+		// Check if instance is shutting down before autostart logic
+		if inst.GetStatus() == instance.ShuttingDown {
+			writeError(w, http.StatusServiceUnavailable, "instance_shutting_down", "Instance is shutting down")
+			return
+		}
+
 		if !inst.IsRemote() && !inst.IsRunning() {
 			err := h.ensureInstanceRunning(inst)
 			if err != nil {
@@ -118,17 +123,16 @@ func (h *Handler) LlamaCppProxy() http.HandlerFunc {
 			}
 		}
 
-		proxy, err := inst.GetProxy()
-		if err != nil {
-			writeError(w, http.StatusInternalServerError, "failed to get proxy", err.Error())
-			return
-		}
-
 		if !inst.IsRemote() {
 			h.stripLlamaCppPrefix(r, inst.Name)
 		}
 
-		proxy.ServeHTTP(w, r)
+		// Use instance's ServeHTTP which tracks inflight requests and handles shutting down state
+		err = inst.ServeHTTP(w, r)
+		if err != nil {
+			// Error is already handled in ServeHTTP (response written)
+			return
+		}
 	}
 }
 

pkg/server/handlers_instances.go

@@ -332,12 +332,6 @@ func (h *Handler) InstanceProxy() http.HandlerFunc {
 			return
 		}
 
-		proxy, err := inst.GetProxy()
-		if err != nil {
-			writeError(w, http.StatusInternalServerError, "proxy_failed", "Failed to get proxy: "+err.Error())
-			return
-		}
-
 		if !inst.IsRemote() {
 			// Strip the "/api/v1/instances/<name>/proxy" prefix from the request URL
 			prefix := fmt.Sprintf("/api/v1/instances/%s/proxy", inst.Name)
@@ -348,6 +342,11 @@ func (h *Handler) InstanceProxy() http.HandlerFunc {
 		r.Header.Set("X-Forwarded-Host", r.Header.Get("Host"))
 		r.Header.Set("X-Forwarded-Proto", "http")
 
-		proxy.ServeHTTP(w, r)
+		// Use instance's ServeHTTP which tracks inflight requests and handles shutting down state
+		err = inst.ServeHTTP(w, r)
+		if err != nil {
+			// Error is already handled in ServeHTTP (response written)
+			return
+		}
 	}
 }

pkg/server/handlers_openai.go

@@ -4,6 +4,7 @@ import (
 	"bytes"
 	"encoding/json"
 	"io"
+	"llamactl/pkg/instance"
 	"llamactl/pkg/validation"
 	"net/http"
 )
@@ -106,6 +107,12 @@ func (h *Handler) OpenAIProxy() http.HandlerFunc {
 			return
 		}
 
+		// Check if instance is shutting down before autostart logic
+		if inst.GetStatus() == instance.ShuttingDown {
+			writeError(w, http.StatusServiceUnavailable, "instance_shutting_down", "Instance is shutting down")
+			return
+		}
+
 		if !inst.IsRemote() && !inst.IsRunning() {
 			err := h.ensureInstanceRunning(inst)
 			if err != nil {
@@ -114,16 +121,15 @@ func (h *Handler) OpenAIProxy() http.HandlerFunc {
 			}
 		}
 
-		proxy, err := inst.GetProxy()
-		if err != nil {
-			writeError(w, http.StatusInternalServerError, "proxy_failed", err.Error())
-			return
-		}
-
 		// Recreate the request body from the bytes we read
 		r.Body = io.NopCloser(bytes.NewReader(bodyBytes))
 		r.ContentLength = int64(len(bodyBytes))
 
-		proxy.ServeHTTP(w, r)
+		// Use instance's ServeHTTP which tracks inflight requests and handles shutting down state
+		err = inst.ServeHTTP(w, r)
+		if err != nil {
+			// Error is already handled in ServeHTTP (response written)
+			return
+		}
 	}
 }

pkg/server/routes.go

@@ -1,7 +1,7 @@
 package server
 
 import (
-	"fmt"
+	"log"
 
 	"github.com/go-chi/chi/v5"
 	"github.com/go-chi/chi/v5/middleware"
@@ -159,7 +159,7 @@ func SetupRouter(handler *Handler) *chi.Mux {
 
 	// Serve WebUI files
 	if err := webui.SetupWebUI(r); err != nil {
-		fmt.Printf("Failed to set up WebUI: %v\n", err)
+		log.Printf("Failed to set up WebUI: %v\n", err)
 	}
 
 	return r

webui/src/components/BackendFormField.tsx

@@ -12,13 +12,13 @@ interface BackendFormFieldProps {
 
 const BackendFormField: React.FC<BackendFormFieldProps> = ({ fieldKey, value, onChange }) => {
   // Get configuration for basic fields, or use field name for advanced fields
-  const config = basicBackendFieldsConfig[fieldKey as string] || { label: fieldKey }
+  const config = basicBackendFieldsConfig[fieldKey] || { label: fieldKey }
   
   // Get type from Zod schema
   const fieldType = getBackendFieldType(fieldKey)
 
   const handleChange = (newValue: string | number | boolean | string[] | undefined) => {
-    onChange(fieldKey as string, newValue)
+    onChange(fieldKey, newValue)
   }
 
   const renderField = () => {

webui/src/components/HealthBadge.tsx

@@ -21,6 +21,8 @@ const HealthBadge: React.FC<HealthBadgeProps> = ({ health }) => {
         return <Loader2 className="h-3 w-3 animate-spin" />;
       case "restarting":
         return <Loader2 className="h-3 w-3 animate-spin" />;
+      case "shutting_down":
+        return <Loader2 className="h-3 w-3 animate-spin" />;
       case "stopped":
         return <Clock className="h-3 w-3" />;
       case "failed":
@@ -36,6 +38,8 @@ const HealthBadge: React.FC<HealthBadgeProps> = ({ health }) => {
         return "outline";
       case "restarting":
         return "outline";
+      case "shutting_down":
+        return "outline";
       case "stopped":
         return "secondary";
       case "failed":
@@ -51,6 +55,8 @@ const HealthBadge: React.FC<HealthBadgeProps> = ({ health }) => {
         return "Starting";
       case "restarting":
         return "Restarting";
+      case "shutting_down":
+        return "Shutting Down";
       case "stopped":
         return "Stopped";
       case "failed":

webui/src/components/InstanceCard.tsx

@@ -2,12 +2,13 @@
 import { Button } from "@/components/ui/button";
 import { Card, CardContent, CardHeader, CardTitle } from "@/components/ui/card";
 import type { Instance } from "@/types/instance";
-import { Edit, FileText, Play, Square, Trash2, MoreHorizontal } from "lucide-react";
+import { Edit, FileText, Play, Square, Trash2, MoreHorizontal, Download } from "lucide-react";
 import LogsDialog from "@/components/LogDialog";
 import HealthBadge from "@/components/HealthBadge";
 import BackendBadge from "@/components/BackendBadge";
 import { useState } from "react";
 import { useInstanceHealth } from "@/hooks/useInstanceHealth";
+import { instancesApi } from "@/lib/api";
 
 interface InstanceCardProps {
   instance: Instance;
@@ -52,6 +53,40 @@ function InstanceCard({
     setIsLogsOpen(true);
   };
 
+  const handleExport = () => {
+    void (async () => {
+      try {
+        // Fetch the most up-to-date instance data from the backend
+        const instanceData = await instancesApi.get(instance.name);
+
+        // Remove docker_enabled as it's a computed field, not persisted to disk
+        // eslint-disable-next-line @typescript-eslint/no-unused-vars
+        const { docker_enabled, ...persistedData } = instanceData;
+
+        // Convert to JSON string with pretty formatting (matching backend format)
+        const jsonString = JSON.stringify(persistedData, null, 2);
+
+        // Create a blob and download link
+        const blob = new Blob([jsonString], { type: "application/json" });
+        const url = URL.createObjectURL(blob);
+        const link = document.createElement("a");
+        link.href = url;
+        link.download = `${instance.name}.json`;
+
+        // Trigger download
+        document.body.appendChild(link);
+        link.click();
+
+        // Cleanup
+        document.body.removeChild(link);
+        URL.revokeObjectURL(url);
+      } catch (error) {
+        console.error("Failed to export instance:", error);
+        alert(`Failed to export instance: ${error instanceof Error ? error.message : "Unknown error"}`);
+      }
+    })();
+  };
+
   const running = instance.status === "running";
 
   return (
@@ -131,6 +166,18 @@ function InstanceCard({
                 Logs
               </Button>
 
+              <Button
+                size="sm"
+                variant="outline"
+                onClick={handleExport}
+                title="Export instance"
+                data-testid="export-instance-button"
+                className="flex-1"
+              >
+                <Download className="h-4 w-4 mr-1" />
+                Export
+              </Button>
+
               <Button
                 size="sm"
                 variant="destructive"

webui/src/components/InstanceDialog.tsx

@@ -1,4 +1,4 @@
-import React, { useState, useEffect } from "react";
+import React, { useState, useEffect, useRef } from "react";
 import { Button } from "@/components/ui/button";
 import {
   Dialog,
@@ -9,9 +9,11 @@ import {
   DialogTitle,
 } from "@/components/ui/dialog";
 import { BackendType, type CreateInstanceOptions, type Instance } from "@/types/instance";
+import type { BackendOptions } from "@/schemas/instanceOptions";
 import ParseCommandDialog from "@/components/ParseCommandDialog";
 import InstanceSettingsCard from "@/components/instance/InstanceSettingsCard";
 import BackendConfigurationCard from "@/components/instance/BackendConfigurationCard";
+import { Upload } from "lucide-react";
 
 interface InstanceDialogProps {
   open: boolean;
@@ -32,6 +34,7 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
   const [formData, setFormData] = useState<CreateInstanceOptions>({});
   const [nameError, setNameError] = useState("");
   const [showParseDialog, setShowParseDialog] = useState(false);
+  const fileInputRef = useRef<HTMLInputElement>(null);
 
 
   // Reset form when dialog opens/closes or when instance changes
@@ -54,13 +57,13 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
     }
   }, [open, instance]);
 
-  const handleFieldChange = (key: keyof CreateInstanceOptions, value: any) => {
+  const handleFieldChange = (key: keyof CreateInstanceOptions, value: unknown) => {
     setFormData((prev) => {
       // If backend_type is changing, clear backend_options
       if (key === 'backend_type' && prev.backend_type !== value) {
         return {
           ...prev,
-          [key]: value,
+          backend_type: value as CreateInstanceOptions['backend_type'],
           backend_options: {}, // Clear backend options when backend type changes
         };
       }
@@ -68,17 +71,17 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
       return {
         ...prev,
         [key]: value,
-      };
+      } as CreateInstanceOptions;
     });
   };
 
-  const handleBackendFieldChange = (key: string, value: any) => {
+  const handleBackendFieldChange = (key: string, value: unknown) => {
     setFormData((prev) => ({
       ...prev,
       backend_options: {
         ...prev.backend_options,
         [key]: value,
-      } as any,
+      } as BackendOptions,
     }));
   };
 
@@ -104,11 +107,13 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
     }
 
     // Clean up undefined values to avoid sending empty fields
-    const cleanOptions: CreateInstanceOptions = {};
+    const cleanOptions: CreateInstanceOptions = {} as CreateInstanceOptions;
     Object.entries(formData).forEach(([key, value]) => {
+      const typedKey = key as keyof CreateInstanceOptions;
+
       if (key === 'backend_options' && value && typeof value === 'object' && !Array.isArray(value)) {
         // Handle backend_options specially - clean nested object
-        const cleanBackendOptions: any = {};
+        const cleanBackendOptions: Record<string, unknown> = {};
         Object.entries(value).forEach(([backendKey, backendValue]) => {
           if (backendValue !== undefined && backendValue !== null && (typeof backendValue !== 'string' || backendValue.trim() !== "")) {
             // Handle arrays - don't include empty arrays
@@ -121,7 +126,7 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
 
         // Only include backend_options if it has content
         if (Object.keys(cleanBackendOptions).length > 0) {
-          (cleanOptions as any)[key] = cleanBackendOptions;
+          (cleanOptions as Record<string, unknown>)[typedKey] = cleanBackendOptions as BackendOptions;
         }
       } else if (value !== undefined && value !== null) {
         // Skip empty strings
@@ -132,7 +137,7 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
         if (Array.isArray(value) && value.length === 0) {
           return;
         }
-        (cleanOptions as any)[key] = value;
+        (cleanOptions as Record<string, unknown>)[typedKey] = value;
       }
     });
 
@@ -153,6 +158,49 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
     setShowParseDialog(false);
   };
 
+  const handleImportFile = () => {
+    fileInputRef.current?.click();
+  };
+
+  const handleFileChange = (event: React.ChangeEvent<HTMLInputElement>) => {
+    const file = event.target.files?.[0];
+    if (!file) return;
+
+    const reader = new FileReader();
+    reader.onload = (e) => {
+      try {
+        const content = e.target?.result as string;
+        const importedData = JSON.parse(content) as { name?: string; options?: CreateInstanceOptions };
+
+        // Validate that it's an instance export
+        if (!importedData.name || !importedData.options) {
+          alert('Invalid instance file: Missing required fields (name, options)');
+          return;
+        }
+
+        // Set the instance name (only for new instances, not editing)
+        if (!isEditing && typeof importedData.name === 'string') {
+          handleNameChange(importedData.name);
+        }
+
+        // Populate all the options from the imported file
+        if (importedData.options) {
+          setFormData(prev => ({
+            ...prev,
+            ...importedData.options,
+          }));
+        }
+
+        // Reset the file input
+        event.target.value = '';
+      } catch (error) {
+        console.error('Failed to parse instance file:', error);
+        alert(`Failed to parse instance file: ${error instanceof Error ? error.message : 'Invalid JSON'}`);
+      }
+    };
+
+    reader.readAsText(file);
+  };
 
   // Save button label logic
   let saveButtonLabel = "Create Instance";
@@ -168,14 +216,38 @@ const InstanceDialog: React.FC<InstanceDialogProps> = ({
     <Dialog open={open} onOpenChange={onOpenChange}>
       <DialogContent className="sm:max-w-[600px] max-h-[80vh] overflow-hidden flex flex-col">
         <DialogHeader>
-          <DialogTitle>
+          <div className="flex items-center justify-between">
-            {isEditing ? "Edit Instance" : "Create New Instance"}
+            <div className="flex-1">
-          </DialogTitle>
+              <DialogTitle>
-          <DialogDescription>
+                {isEditing ? "Edit Instance" : "Create New Instance"}
-            {isEditing
+              </DialogTitle>
-              ? "Modify the instance configuration below."
+              <DialogDescription>
-              : "Configure your new llama-server instance below."}
+                {isEditing
-          </DialogDescription>
+                  ? "Modify the instance configuration below."
+                  : "Configure your new llama-server instance below."}
+              </DialogDescription>
+            </div>
+            {!isEditing && (
+              <Button
+                type="button"
+                variant="ghost"
+                size="sm"
+                onClick={handleImportFile}
+                title="Import instance configuration from JSON file"
+                className="ml-2"
+              >
+                <Upload className="h-4 w-4 mr-2" />
+                Import
+              </Button>
+            )}
+          </div>
+          <input
+            ref={fileInputRef}
+            type="file"
+            accept=".json"
+            onChange={handleFileChange}
+            className="hidden"
+          />
         </DialogHeader>
 
         <div className="flex-1 overflow-y-auto">

webui/src/components/InstanceList.tsx

@@ -56,9 +56,9 @@ function InstanceList({ editInstance }: InstanceListProps) {
           <MemoizedInstanceCard 
             key={instance.name} 
             instance={instance}
-            startInstance={startInstance}
+            startInstance={() => { void startInstance(instance.name) }}
-            stopInstance={stopInstance}
+            stopInstance={() => { void stopInstance(instance.name) }}
-            deleteInstance={deleteInstance}
+            deleteInstance={() => { void deleteInstance(instance.name) }}
             editInstance={editInstance}
           />
         ))}

webui/src/components/ParseCommandDialog.tsx

@@ -54,7 +54,7 @@ const ParseCommandDialog: React.FC<ParseCommandDialogProps> = ({
           options = await backendsApi.vllm.parseCommand(command);
           break;
         default:
-          throw new Error(`Unsupported backend type: ${backendType}`);
+          throw new Error(`Unsupported backend type: ${String(backendType)}`);
       }
 
       onParsed(options);

webui/src/hooks/useInstanceHealth.ts

@@ -18,7 +18,7 @@ export function useInstanceHealth(instanceName: string, instanceStatus: Instance
 
   // Trigger health check when instance status changes to active states
   useEffect(() => {
-    if (instanceStatus === 'running' || instanceStatus === 'restarting') {
+    if (instanceStatus === 'running' || instanceStatus === 'restarting' || instanceStatus === 'shutting_down') {
       healthService.refreshHealth(instanceName).catch(error => {
         console.error(`Failed to refresh health for ${instanceName}:`, error)
       })

webui/src/lib/healthService.ts

@@ -5,11 +5,12 @@ type HealthCallback = (health: HealthStatus) => void
 
 // Polling intervals based on health state (in milliseconds)
 const POLLING_INTERVALS: Record<HealthState, number> = {
-  'starting': 5000,    // 5 seconds - frequent during startup
+  'starting': 5000,      // 5 seconds - frequent during startup
-  'restarting': 5000,  // 5 seconds - restart in progress
+  'restarting': 5000,    // 5 seconds - restart in progress
-  'ready': 60000,      // 60 seconds - stable state
+  'shutting_down': 3000, // 3 seconds - monitor shutdown progress
-  'stopped': 0,        // No polling
+  'ready': 60000,        // 60 seconds - stable state
-  'failed': 0,         // No polling
+  'stopped': 0,          // No polling
+  'failed': 0,           // No polling
 }
 
 class HealthService {
@@ -96,6 +97,7 @@ class HealthService {
       case 'running': return 'starting' // Should not happen as we check HTTP for running
       case 'failed': return 'failed'
       case 'restarting': return 'restarting'
+      case 'shutting_down': return 'shutting_down'
     }
   }
 

webui/src/types/instance.ts

@@ -11,9 +11,9 @@ export const BackendType = {
 
 export type BackendTypeValue = typeof BackendType[keyof typeof BackendType]
 
-export type InstanceStatus = 'running' | 'stopped' | 'failed' | 'restarting'
+export type InstanceStatus = 'running' | 'stopped' | 'failed' | 'restarting' | 'shutting_down'
 
-export type HealthState = 'stopped' | 'starting' | 'ready' | 'failed' | 'restarting'
+export type HealthState = 'stopped' | 'starting' | 'ready' | 'failed' | 'restarting' | 'shutting_down'
 
 export interface HealthStatus {
   state: HealthState