Refactor instance status management: replace Running boolean with InstanceStatus enum and update related methods

pkg/instance/instance.go

@@ -82,7 +82,7 @@ type Process struct {
 	globalSettings *config.InstancesConfig
 
 	// Status
-	Running bool `json:"running"`
+	Status InstanceStatus `json:"status"`
 
 	// Creation time
 	Created int64 `json:"created,omitempty"` // Unix timestamp when the instance was created
@@ -287,12 +287,12 @@ func (i *Process) MarshalJSON() ([]byte, error) {
 	temp := struct {
 		Name    string                 `json:"name"`
 		Options *CreateInstanceOptions `json:"options,omitempty"`
-		Running bool                   `json:"running"`
+		Status  InstanceStatus         `json:"status"`
 		Created int64                  `json:"created,omitempty"`
 	}{
 		Name:    i.Name,
 		Options: i.options,
-		Running: i.Running,
+		Status:  i.Status,
 		Created: i.Created,
 	}
 
@@ -305,7 +305,7 @@ func (i *Process) UnmarshalJSON(data []byte) error {
 	temp := struct {
 		Name    string                 `json:"name"`
 		Options *CreateInstanceOptions `json:"options,omitempty"`
-		Running bool                   `json:"running"`
+		Status  InstanceStatus         `json:"status"`
 		Created int64                  `json:"created,omitempty"`
 	}{}
 
@@ -315,7 +315,7 @@ func (i *Process) UnmarshalJSON(data []byte) error {
 
 	// Set the fields
 	i.Name = temp.Name
-	i.Running = temp.Running
+	i.Status = temp.Status
 	i.Created = temp.Created
 
 	// Handle options with validation but no defaults

pkg/instance/instance_test.go

@@ -29,7 +29,7 @@ func TestNewInstance(t *testing.T) {
 	if instance.Name != "test-instance" {
 		t.Errorf("Expected name 'test-instance', got %q", instance.Name)
 	}
-	if instance.Running {
+	if instance.IsRunning() {
 		t.Error("New instance should not be running")
 	}
 
@@ -188,7 +188,7 @@ func TestMarshalJSON(t *testing.T) {
 	}
 
 	// Check that JSON contains expected fields
-	var result map[string]interface{}
+	var result map[string]any
 	err = json.Unmarshal(data, &result)
 	if err != nil {
 		t.Fatalf("JSON unmarshal failed: %v", err)
@@ -197,8 +197,8 @@ func TestMarshalJSON(t *testing.T) {
 	if result["name"] != "test-instance" {
 		t.Errorf("Expected name 'test-instance', got %v", result["name"])
 	}
-	if result["running"] != false {
+	if result["status"] != "stopped" {
-		t.Errorf("Expected running false, got %v", result["running"])
+		t.Errorf("Expected status 'stopped', got %v", result["status"])
 	}
 
 	// Check that options are included
@@ -218,7 +218,7 @@ func TestMarshalJSON(t *testing.T) {
 func TestUnmarshalJSON(t *testing.T) {
 	jsonData := `{
 		"name": "test-instance",
-		"running": true,
+		"status": "running",
 		"options": {
 			"model": "/path/to/model.gguf",
 			"port": 8080,
@@ -236,8 +236,8 @@ func TestUnmarshalJSON(t *testing.T) {
 	if inst.Name != "test-instance" {
 		t.Errorf("Expected name 'test-instance', got %q", inst.Name)
 	}
-	if !inst.Running {
+	if !inst.IsRunning() {
-		t.Error("Expected running to be true")
+		t.Error("Expected status to be running")
 	}
 
 	opts := inst.GetOptions()

pkg/instance/lifecycle.go

@@ -11,18 +11,12 @@ import (
 	"time"
 )
 
-func (i *Process) IsRunning() bool {
-	i.mu.RLock()
-	defer i.mu.RUnlock()
-	return i.Running
-}
-
 // Start starts the llama server instance and returns an error if it fails.
 func (i *Process) Start() error {
 	i.mu.Lock()
 	defer i.mu.Unlock()
 
-	if i.Running {
+	if i.IsRunning() {
 		return fmt.Errorf("instance %s is already running", i.Name)
 	}
 
@@ -71,7 +65,7 @@ func (i *Process) Start() error {
 		return fmt.Errorf("failed to start instance %s: %w", i.Name, err)
 	}
 
-	i.Running = true
+	i.SetStatus(Running)
 
 	// Create channel for monitor completion signaling
 	i.monitorDone = make(chan struct{})
@@ -88,7 +82,7 @@ func (i *Process) Start() error {
 func (i *Process) Stop() error {
 	i.mu.Lock()
 
-	if !i.Running {
+	if !i.IsRunning() {
 		// Even if not running, cancel any pending restart
 		if i.restartCancel != nil {
 			i.restartCancel()
@@ -105,8 +99,8 @@ func (i *Process) Stop() error {
 		i.restartCancel = nil
 	}
 
-	// Set running to false first to signal intentional stop
+	// Set status to stopped first to signal intentional stop
-	i.Running = false
+	i.SetStatus(Stopped)
 
 	// Clean up the proxy
 	i.proxy = nil
@@ -151,7 +145,7 @@ func (i *Process) Stop() error {
 }
 
 func (i *Process) WaitForHealthy(timeout int) error {
-	if !i.Running {
+	if !i.IsRunning() {
 		return fmt.Errorf("instance %s is not running", i.Name)
 	}
 
@@ -233,12 +227,12 @@ func (i *Process) monitorProcess() {
 	i.mu.Lock()
 
 	// Check if the instance was intentionally stopped
-	if !i.Running {
+	if !i.IsRunning() {
 		i.mu.Unlock()
 		return
 	}
 
-	i.Running = false
+	i.SetStatus(Stopped)
 	i.logger.Close()
 
 	// Cancel any existing restart context since we're handling a new exit

pkg/instance/status.go

@@ -0,0 +1,72 @@
+package instance
+
+import (
+	"encoding/json"
+	"log"
+)
+
+// Enum for instance status
+type InstanceStatus int
+
+const (
+	Stopped InstanceStatus = iota
+	Running
+	Failed
+)
+
+var nameToStatus = map[string]InstanceStatus{
+	"stopped": Stopped,
+	"running": Running,
+	"failed":  Failed,
+}
+
+var statusToName = map[InstanceStatus]string{
+	Stopped: "stopped",
+	Running: "running",
+	Failed:  "failed",
+}
+
+func (p *Process) SetStatus(status InstanceStatus) {
+	p.mu.Lock()
+	defer p.mu.Unlock()
+
+	p.Status = status
+}
+
+func (p *Process) GetStatus() InstanceStatus {
+	p.mu.RLock()
+	defer p.mu.RUnlock()
+	return p.Status
+}
+
+// IsRunning returns true if the status is Running
+func (p *Process) IsRunning() bool {
+	p.mu.RLock()
+	defer p.mu.RUnlock()
+	return p.Status == Running
+}
+
+func (s InstanceStatus) MarshalJSON() ([]byte, error) {
+	name, ok := statusToName[s]
+	if !ok {
+		name = "stopped" // Default to "stopped" for unknown status
+	}
+	return json.Marshal(name)
+}
+
+// UnmarshalJSON implements json.Unmarshaler
+func (s *InstanceStatus) UnmarshalJSON(data []byte) error {
+	var str string
+	if err := json.Unmarshal(data, &str); err != nil {
+		return err
+	}
+
+	status, ok := nameToStatus[str]
+	if !ok {
+		log.Printf("Unknown instance status: %s", str)
+		status = Stopped // Default to Stopped on unknown status
+	}
+
+	*s = status
+	return nil
+}

pkg/instance/timeout.go

@@ -13,7 +13,7 @@ func (i *Process) ShouldTimeout() bool {
 	i.mu.RLock()
 	defer i.mu.RUnlock()
 
-	if !i.Running || i.options.IdleTimeout == nil || *i.options.IdleTimeout <= 0 {
+	if !i.IsRunning() || i.options.IdleTimeout == nil || *i.options.IdleTimeout <= 0 {
 		return false
 	}
 

pkg/instance/timeout_test.go

@@ -94,7 +94,7 @@ func TestShouldTimeout_NoTimeoutConfigured(t *testing.T) {
 
 			inst := instance.NewInstance("test-instance", globalSettings, options)
 			// Simulate running state
-			inst.Running = true
+			inst.SetStatus(instance.Running)
 
 			if inst.ShouldTimeout() {
 				t.Errorf("Instance with %s should not timeout", tt.name)
@@ -117,7 +117,7 @@ func TestShouldTimeout_WithinTimeLimit(t *testing.T) {
 	}
 
 	inst := instance.NewInstance("test-instance", globalSettings, options)
-	inst.Running = true
+	inst.SetStatus(instance.Running)
 
 	// Update last request time to now
 	inst.UpdateLastRequestTime()
@@ -142,7 +142,7 @@ func TestShouldTimeout_ExceedsTimeLimit(t *testing.T) {
 	}
 
 	inst := instance.NewInstance("test-instance", globalSettings, options)
-	inst.Running = true
+	inst.SetStatus(instance.Running)
 
 	// Use MockTimeProvider to simulate old last request time
 	mockTime := NewMockTimeProvider(time.Now())

pkg/manager/manager.go

@@ -150,7 +150,7 @@ func (im *instanceManager) Shutdown() {
 	wg.Add(len(im.instances))
 
 	for name, inst := range im.instances {
-		if !inst.Running {
+		if !inst.IsRunning() {
 			wg.Done() // If instance is not running, just mark it as done
 			continue
 		}
@@ -234,7 +234,7 @@ func (im *instanceManager) loadInstance(name, path string) error {
 
 	// Restore persisted fields that NewInstance doesn't set
 	inst.Created = persistedInstance.Created
-	inst.Running = persistedInstance.Running
+	inst.SetStatus(persistedInstance.Status)
 
 	// Check for port conflicts and add to maps
 	if inst.GetOptions() != nil && inst.GetOptions().Port > 0 {
@@ -254,7 +254,7 @@ func (im *instanceManager) autoStartInstances() {
 	im.mu.RLock()
 	var instancesToStart []*instance.Process
 	for _, inst := range im.instances {
-		if inst.Running && // Was running when persisted
+		if inst.IsRunning() && // Was running when persisted
 			inst.GetOptions() != nil &&
 			inst.GetOptions().AutoRestart != nil &&
 			*inst.GetOptions().AutoRestart {
@@ -266,7 +266,7 @@ func (im *instanceManager) autoStartInstances() {
 	for _, inst := range instancesToStart {
 		log.Printf("Auto-starting instance %s", inst.Name)
 		// Reset running state before starting (since Start() expects stopped instance)
-		inst.Running = false
+		inst.SetStatus(instance.Stopped)
 		if err := inst.Start(); err != nil {
 			log.Printf("Failed to auto-start instance %s: %v", inst.Name, err)
 		}

pkg/manager/manager_test.go

@@ -59,7 +59,7 @@ func TestCreateInstance_Success(t *testing.T) {
 	if inst.Name != "test-instance" {
 		t.Errorf("Expected instance name 'test-instance', got %q", inst.Name)
 	}
-	if inst.Running {
+	if inst.GetStatus() != instance.Stopped {
 		t.Error("New instance should not be running")
 	}
 	if inst.GetOptions().Port != 8080 {
@@ -357,7 +357,7 @@ func TestTimeoutFunctionality(t *testing.T) {
 	inst.SetTimeProvider(mockTime)
 
 	// Set instance to running state so timeout logic can work
-	inst.Running = true
+	inst.SetStatus(instance.Running)
 
 	// Simulate instance being "running" for timeout check (without actual process)
 	// We'll test the ShouldTimeout logic directly
@@ -377,7 +377,7 @@ func TestTimeoutFunctionality(t *testing.T) {
 	}
 
 	// Reset running state to avoid shutdown issues
-	inst.Running = false
+	inst.SetStatus(instance.Stopped)
 
 	// Test that instance without timeout doesn't timeout
 	noTimeoutOptions := &instance.CreateInstanceOptions{
@@ -393,7 +393,7 @@ func TestTimeoutFunctionality(t *testing.T) {
 	}
 
 	noTimeoutInst.SetTimeProvider(mockTime)
-	noTimeoutInst.Running = true // Set to running for timeout check
+	noTimeoutInst.SetStatus(instance.Running) // Set to running for timeout check
 	noTimeoutInst.UpdateLastRequestTime()
 
 	// Even with time advanced, should not timeout
@@ -402,7 +402,7 @@ func TestTimeoutFunctionality(t *testing.T) {
 	}
 
 	// Reset running state to avoid shutdown issues
-	noTimeoutInst.Running = false
+	noTimeoutInst.SetStatus(instance.Stopped)
 }
 
 func TestConcurrentAccess(t *testing.T) {

pkg/manager/operations.go

@@ -109,7 +109,7 @@ func (im *instanceManager) UpdateInstance(name string, options *instance.CreateI
 	}
 
 	// Check if instance is running before updating options
-	wasRunning := instance.Running
+	wasRunning := instance.IsRunning()
 
 	// If the instance is running, stop it first
 	if wasRunning {
@@ -147,7 +147,7 @@ func (im *instanceManager) DeleteInstance(name string) error {
 		return fmt.Errorf("instance with name %s not found", name)
 	}
 
-	if instance.Running {
+	if instance.IsRunning() {
 		return fmt.Errorf("instance with name %s is still running, stop it before deleting", name)
 	}
 
@@ -173,7 +173,7 @@ func (im *instanceManager) StartInstance(name string) (*instance.Process, error)
 	if !exists {
 		return nil, fmt.Errorf("instance with name %s not found", name)
 	}
-	if instance.Running {
+	if instance.IsRunning() {
 		return instance, fmt.Errorf("instance with name %s is already running", name)
 	}
 
@@ -200,7 +200,7 @@ func (im *instanceManager) StopInstance(name string) (*instance.Process, error)
 	if !exists {
 		return nil, fmt.Errorf("instance with name %s not found", name)
 	}
-	if !instance.Running {
+	if !instance.IsRunning() {
 		return instance, fmt.Errorf("instance with name %s is already stopped", name)
 	}
 

pkg/server/handlers.go

@@ -451,7 +451,7 @@ func (h *Handler) ProxyToInstance() http.HandlerFunc {
 			return
 		}
 
-		if !inst.Running {
+		if !inst.IsRunning() {
 			http.Error(w, "Instance is not running", http.StatusServiceUnavailable)
 			return
 		}
@@ -574,7 +574,7 @@ func (h *Handler) OpenAIProxy() http.HandlerFunc {
 			return
 		}
 
-		if !inst.Running {
+		if !inst.IsRunning() {
 			if inst.GetOptions().OnDemandStart != nil && *inst.GetOptions().OnDemandStart {
 				// If on-demand start is enabled, start the instance
 				if _, err := h.InstanceManager.StartInstance(modelName); err != nil {