| 1 | package main |
| 2 | |
| 3 | import ( |
| 4 | "bufio" |
| 5 | "context" |
| 6 | "crypto/rand" |
| 7 | "encoding/base64" |
| 8 | "encoding/json" |
| 9 | "fmt" |
| 10 | "log" |
| 11 | "math" |
| 12 | "math/big" |
| 13 | "net/http" |
| 14 | "os" |
| 15 | "os/signal" |
| 16 | "strings" |
| 17 | "sync" |
| 18 | "syscall" |
| 19 | "time" |
| 20 | |
| 21 | "github.com/coder/websocket" |
| 22 | ) |
| 23 | |
| 24 | // Connection states |
| 25 | const ( |
| 26 | stateDisconnected = iota |
| 27 | stateConnecting |
| 28 | stateConnected |
| 29 | ) |
| 30 | |
| 31 | // WebSocket message types (these are defined in the websocket package but we need to define them here) |
| 32 | const ( |
| 33 | MessageText = websocket.MessageText |
| 34 | MessageBinary = websocket.MessageBinary |
| 35 | ) |
| 36 | |
| 37 | // JSON-RPC 2.0 related structures |
| 38 | type JsonRpcMessage struct { |
| 39 | JsonRpc string `json:"jsonrpc"` |
| 40 | Id interface{} `json:"id,omitempty"` |
| 41 | Method string `json:"method,omitempty"` |
| 42 | Result interface{} `json:"result,omitempty"` |
| 43 | Error *JsonRpcError `json:"error,omitempty"` |
| 44 | } |
| 45 | |
| 46 | type JsonRpcError struct { |
| 47 | Code int `json:"code"` |
| 48 | Message string `json:"message"` |
| 49 | Data interface{} `json:"data,omitempty"` |
| 50 | } |
| 51 | |
| 52 | // ConnectionTimeout is the timeout for initial connection attempts |
| 53 | const ConnectionTimeout = 5 * time.Second |
| 54 | |
| 55 | // PendingRequest represents a message waiting for the connection to be established |
| 56 | type PendingRequest struct { |
| 57 | ID interface{} |
| 58 | Message string |
| 59 | Timer *time.Timer |
| 60 | } |
| 61 | |
| 62 | func generateWebSocketKey() string { |
| 63 | key := make([]byte, 16) |
| 64 | _, err := rand.Read(key) |
| 65 | if err != nil { |
| 66 | log.Fatalf("failed to generate WebSocket key: %v", err) |
| 67 | } |
| 68 | return base64.StdEncoding.EncodeToString(key) |
| 69 | } |
| 70 | |
| 71 | func main() { |
| 72 | // Get program name for logs |
| 73 | programName := "nd-mcp-golang" |
| 74 | if len(os.Args) > 0 { |
| 75 | programName = os.Args[0] |
| 76 | } |
| 77 | |
| 78 | args := os.Args[1:] |
| 79 | var targetURL string |
| 80 | var bearerToken string |
| 81 | |
| 82 | for len(args) > 0 { |
| 83 | arg := args[0] |
| 84 | switch { |
| 85 | case arg == "--bearer": |
| 86 | if len(args) < 2 { |
| 87 | fmt.Fprintf(os.Stderr, "%s: Usage: %s [--bearer TOKEN] ws://host/path\n", programName, programName) |
| 88 | os.Exit(1) |
| 89 | } |
| 90 | bearerToken = strings.TrimSpace(args[1]) |
| 91 | args = args[2:] |
| 92 | case strings.HasPrefix(arg, "--bearer="): |
| 93 | bearerToken = strings.TrimSpace(strings.TrimPrefix(arg, "--bearer=")) |
| 94 | args = args[1:] |
| 95 | default: |
| 96 | if targetURL != "" { |
| 97 | fmt.Fprintf(os.Stderr, "%s: Unexpected argument '%s'\n", programName, arg) |
| 98 | fmt.Fprintf(os.Stderr, "%s: Usage: %s [--bearer TOKEN] ws://host/path\n", programName, programName) |
| 99 | os.Exit(1) |
| 100 | } |
| 101 | targetURL = arg |
| 102 | args = args[1:] |
| 103 | } |
| 104 | } |
| 105 | |
| 106 | if targetURL == "" { |
| 107 | fmt.Fprintf(os.Stderr, "%s: Usage: %s [--bearer TOKEN] ws://host/path\n", programName, programName) |
| 108 | os.Exit(1) |
| 109 | } |
| 110 | |
| 111 | if bearerToken == "" { |
| 112 | bearerToken = strings.TrimSpace(os.Getenv("ND_MCP_BEARER_TOKEN")) |
| 113 | } |
| 114 | |
| 115 | if bearerToken != "" { |
| 116 | fmt.Fprintf(os.Stderr, "%s: Authorization header enabled for MCP connection\n", programName) |
| 117 | } |
| 118 | |
| 119 | // Set up channels for communication |
| 120 | stdinCh := make(chan string, 100) // Buffer stdin messages |
| 121 | reconnectCh := make(chan struct{}, 1) // Signal for immediate reconnection |
| 122 | doneCh := make(chan struct{}) // Signal for program termination |
| 123 | stdinClosedCh := make(chan struct{}) // Signal that stdin is closed |
| 124 | |
| 125 | // Global state |
| 126 | var state int // Connection state |
| 127 | var stateMu sync.Mutex |
| 128 | var messageQueueMu sync.Mutex |
| 129 | messageQueue := []string{} |
| 130 | stdinActive := true |
| 131 | |
| 132 | // Pending requests that are waiting for connection to be established |
| 133 | var pendingMu sync.Mutex |
| 134 | pendingRequests := make(map[interface{}]*PendingRequest) |
| 135 | |
| 136 | // Parse a JSON-RPC message and extract the ID and method |
| 137 | parseJsonRpcMessage := func(message string) (interface{}, string) { |
| 138 | var msg JsonRpcMessage |
| 139 | err := json.Unmarshal([]byte(message), &msg) |
| 140 | if err != nil || msg.JsonRpc != "2.0" { |
| 141 | return nil, "" |
| 142 | } |
| 143 | return msg.Id, msg.Method |
| 144 | } |
| 145 | |
| 146 | // Create a JSON-RPC error response |
| 147 | createJsonRpcError := func(id interface{}, code int, message string, data interface{}) string { |
| 148 | response := JsonRpcMessage{ |
| 149 | JsonRpc: "2.0", |
| 150 | Id: id, |
| 151 | Error: &JsonRpcError{ |
| 152 | Code: code, |
| 153 | Message: message, |
| 154 | Data: data, |
| 155 | }, |
| 156 | } |
| 157 | responseJson, err := json.Marshal(response) |
| 158 | if err != nil { |
| 159 | fmt.Fprintf(os.Stderr, "%s: ERROR: Failed to marshal error response: %v\n", programName, err) |
| 160 | return fmt.Sprintf("{\"jsonrpc\":\"2.0\",\"id\":%v,\"error\":{\"code\":%d,\"message\":\"%s\"}}", id, code, message) |
| 161 | } |
| 162 | return string(responseJson) |
| 163 | } |
| 164 | |
| 165 | // Handle a request timeout for connection establishment ONLY |
| 166 | handleRequestTimeout := func(msgId interface{}) { |
| 167 | stateMu.Lock() |
| 168 | currentState := state |
| 169 | stateMu.Unlock() |
| 170 | |
| 171 | // This timeout ONLY applies if the connection is not established yet |
| 172 | if currentState != stateConnected { |
| 173 | pendingMu.Lock() |
| 174 | if req, exists := pendingRequests[msgId]; exists { |
| 175 | fmt.Fprintf(os.Stderr, "%s: Connection timeout for request ID %v, sending error response\n", programName, msgId) |
| 176 | |
| 177 | // Create and send error response |
| 178 | errorResponse := createJsonRpcError( |
| 179 | msgId, |
| 180 | -32000, // Server error code |
| 181 | "MCP server connection failed", |
| 182 | map[string]string{"details": "Could not establish connection to Netdata within timeout period"}, |
| 183 | ) |
| 184 | |
| 185 | fmt.Println(errorResponse) |
| 186 | |
| 187 | // Get the original message |
| 188 | originalMessage := req.Message |
| 189 | |
| 190 | // Remove request from pending |
| 191 | delete(pendingRequests, msgId) |
| 192 | pendingMu.Unlock() |
| 193 | |
| 194 | // Also remove from messageQueue if it exists there |
| 195 | messageQueueMu.Lock() |
| 196 | for i, msg := range messageQueue { |
| 197 | if msg == originalMessage { |
| 198 | // Remove from queue by creating a new slice without this element |
| 199 | messageQueue = append(messageQueue[:i], messageQueue[i+1:]...) |
| 200 | fmt.Fprintf(os.Stderr, "%s: Removed timed-out request from message queue\n", programName) |
| 201 | break |
| 202 | } |
| 203 | } |
| 204 | messageQueueMu.Unlock() |
| 205 | return |
| 206 | } |
| 207 | pendingMu.Unlock() |
| 208 | } else { |
| 209 | // The connection was established before the timeout - just clean up |
| 210 | pendingMu.Lock() |
| 211 | if _, exists := pendingRequests[msgId]; exists { |
| 212 | delete(pendingRequests, msgId) |
| 213 | } |
| 214 | pendingMu.Unlock() |
| 215 | } |
| 216 | } |
| 217 | |
| 218 | // Set up signal handling |
| 219 | sigCh := make(chan os.Signal, 1) |
| 220 | signal.Notify(sigCh, syscall.SIGINT, syscall.SIGTERM) |
| 221 | go func() { |
| 222 | sig := <-sigCh |
| 223 | fmt.Fprintf(os.Stderr, "%s: Received signal %v, shutting down\n", programName, sig) |
| 224 | close(doneCh) |
| 225 | }() |
| 226 | |
| 227 | // Start reading from stdin in a separate goroutine |
| 228 | go func() { |
| 229 | scanner := bufio.NewScanner(os.Stdin) |
| 230 | scannerRunning := true |
| 231 | |
| 232 | // Make scanner buffer larger to handle large messages |
| 233 | const maxScanBufferSize = 1024 * 1024 // 1MB |
| 234 | buf := make([]byte, maxScanBufferSize) |
| 235 | scanner.Buffer(buf, maxScanBufferSize) |
| 236 | |
| 237 | for scannerRunning && scanner.Scan() { |
| 238 | text := scanner.Text() |
| 239 | |
| 240 | // Parse for JSON-RPC ID |
| 241 | msgId, _ := parseJsonRpcMessage(text) |
| 242 | |
| 243 | // Check connection state |
| 244 | stateMu.Lock() |
| 245 | currentState := state |
| 246 | stateMu.Unlock() |
| 247 | |
| 248 | if currentState != stateConnected && msgId != nil { |
| 249 | // Store as pending request with timeout |
| 250 | pendingMu.Lock() |
| 251 | req := &PendingRequest{ |
| 252 | ID: msgId, |
| 253 | Message: text, |
| 254 | Timer: time.AfterFunc(ConnectionTimeout, func() { handleRequestTimeout(msgId) }), |
| 255 | } |
| 256 | pendingRequests[msgId] = req |
| 257 | pendingMu.Unlock() |
| 258 | |
| 259 | fmt.Fprintf(os.Stderr, "%s: Received request with ID %v, setting response timeout\n", programName, msgId) |
| 260 | } |
| 261 | |
| 262 | // Queue the message |
| 263 | stdinCh <- text |
| 264 | |
| 265 | // If we're not connected, trigger immediate reconnection |
| 266 | if currentState != stateConnected { |
| 267 | select { |
| 268 | case reconnectCh <- struct{}{}: |
| 269 | fmt.Fprintf(os.Stderr, "%s: Received stdin data, attempting immediate reconnection\n", programName) |
| 270 | default: |
| 271 | // Channel already has reconnection request, ignore |
| 272 | } |
| 273 | } |
| 274 | } |
| 275 | |
| 276 | // Check for scanner error |
| 277 | if err := scanner.Err(); err != nil { |
| 278 | fmt.Fprintf(os.Stderr, "%s: ERROR: stdin read error: %v\n", programName, err) |
| 279 | } else { |
| 280 | fmt.Fprintf(os.Stderr, "%s: End of stdin\n", programName) |
| 281 | } |
| 282 | |
| 283 | // Signal that stdin is closed |
| 284 | stdinActive = false |
| 285 | close(stdinClosedCh) |
| 286 | }() |
| 287 | |
| 288 | // Reconnection parameters |
| 289 | baseDelay := 1 * time.Second |
| 290 | maxDelay := 60 * time.Second |
| 291 | attempt := 0 |
| 292 | |
| 293 | // Timer for reconnection backoff |
| 294 | var timer *time.Timer |
| 295 | |
| 296 | // Main connection loop |
| 297 | for { |
| 298 | select { |
| 299 | case <-doneCh: |
| 300 | // Program termination requested |
| 301 | return |
| 302 | case <-stdinClosedCh: |
| 303 | // Stdin closed, continue running until websocket disconnects |
| 304 | // and then exit on the next reconnection attempt |
| 305 | stdinClosedCh = nil // Prevent duplicate handling |
| 306 | case <-reconnectCh: |
| 307 | // Immediate reconnection requested (e.g. from stdin activity) |
| 308 | if timer != nil { |
| 309 | timer.Stop() |
| 310 | } |
| 311 | |
| 312 | // Only proceed with immediate reconnection if we're not already connecting/connected |
| 313 | stateMu.Lock() |
| 314 | currentState := state |
| 315 | stateMu.Unlock() |
| 316 | |
| 317 | if currentState == stateDisconnected { |
| 318 | // Reset the reconnection timer |
| 319 | attempt = 0 |
| 320 | // Fall through to connection attempt |
| 321 | } else { |
| 322 | // Already connecting or connected |
| 323 | continue |
| 324 | } |
| 325 | default: |
| 326 | // Calculate backoff delay with jitter |
| 327 | if attempt > 0 { |
| 328 | // Check if stdin is closed and we're disconnected - if so, exit |
| 329 | if !stdinActive { |
| 330 | fmt.Fprintf(os.Stderr, "%s: Stdin closed and disconnected, exiting\n", programName) |
| 331 | return |
| 332 | } |
| 333 | |
| 334 | delaySeconds := math.Min(float64(maxDelay.Seconds()), |
| 335 | float64(baseDelay.Seconds())*math.Pow(2, float64(attempt-1))*(0.5+jitter())) |
| 336 | delay := time.Duration(delaySeconds * float64(time.Second)) |
| 337 | |
| 338 | fmt.Fprintf(os.Stderr, "%s: Reconnecting in %.1f seconds (attempt %d)...\n", |
| 339 | programName, delaySeconds, attempt) |
| 340 | |
| 341 | // Create timer and wait for it to expire, or for signals |
| 342 | timer = time.NewTimer(delay) |
| 343 | |
| 344 | select { |
| 345 | case <-timer.C: |
| 346 | // Timer expired, continue to connection attempt |
| 347 | case <-reconnectCh: |
| 348 | // Immediate reconnection requested |
| 349 | timer.Stop() |
| 350 | case <-doneCh: |
| 351 | // Program termination requested |
| 352 | timer.Stop() |
| 353 | return |
| 354 | case <-stdinClosedCh: |
| 355 | // Stdin closed |
| 356 | timer.Stop() |
| 357 | stdinClosedCh = nil |
| 358 | continue |
| 359 | } |
| 360 | } |
| 361 | } |
| 362 | |
| 363 | // Update state to connecting |
| 364 | stateMu.Lock() |
| 365 | state = stateConnecting |
| 366 | stateMu.Unlock() |
| 367 | |
| 368 | // Set up connection context with cancellation |
| 369 | ctx, cancel := context.WithCancel(context.Background()) |
| 370 | |
| 371 | // Set up connection timeout |
| 372 | connectionCtx, connectionCancel := context.WithTimeout(ctx, 15*time.Second) |
| 373 | defer connectionCancel() |
| 374 | |
| 375 | fmt.Fprintf(os.Stderr, "%s: Connecting to %s...\n", programName, targetURL) |
| 376 | |
| 377 | // Create a custom header with the WebSocket key |
| 378 | header := http.Header{} |
| 379 | header.Set("Sec-WebSocket-Key", generateWebSocketKey()) |
| 380 | header.Set("Sec-WebSocket-Version", "13") |
| 381 | if bearerToken != "" { |
| 382 | header.Set("Authorization", "Bearer "+bearerToken) |
| 383 | } |
| 384 | |
| 385 | // Connect to WebSocket |
| 386 | conn, _, err := websocket.Dial(connectionCtx, targetURL, &websocket.DialOptions{ |
| 387 | CompressionMode: websocket.CompressionContextTakeover, |
| 388 | HTTPHeader: header, |
| 389 | }) |
| 390 | |
| 391 | // Connection failed |
| 392 | if err != nil { |
| 393 | fmt.Fprintf(os.Stderr, "%s: ERROR: websocket connection failed: %v\n", programName, err) |
| 394 | cancel() |
| 395 | |
| 396 | // Increment attempt counter and try again |
| 397 | attempt++ |
| 398 | |
| 399 | // Update state to disconnected |
| 400 | stateMu.Lock() |
| 401 | state = stateDisconnected |
| 402 | stateMu.Unlock() |
| 403 | continue |
| 404 | } |
| 405 | |
| 406 | // Increase the read limit to match Netdata's WS_MAX_INCOMING_FRAME_SIZE (16MB) |
| 407 | conn.SetReadLimit(16 * 1024 * 1024) |
| 408 | |
| 409 | // Update state to connected |
| 410 | stateMu.Lock() |
| 411 | state = stateConnected |
| 412 | stateMu.Unlock() |
| 413 | |
| 414 | fmt.Fprintf(os.Stderr, "%s: Connected\n", programName) |
| 415 | attempt = 0 // Reset attempt counter on successful connection |
| 416 | |
| 417 | // Process any queued messages, including pending requests |
| 418 | messageQueueMu.Lock() |
| 419 | if len(messageQueue) > 0 { |
| 420 | fmt.Fprintf(os.Stderr, "%s: Sending %d queued message(s)\n", programName, len(messageQueue)) |
| 421 | for _, msg := range messageQueue { |
| 422 | if err := conn.Write(ctx, websocket.MessageText, []byte(msg)); err != nil { |
| 423 | fmt.Fprintf(os.Stderr, "%s: ERROR: failed to send queued message: %v\n", programName, err) |
| 424 | } else { |
| 425 | // Check if this was a pending request with an ID |
| 426 | msgId, _ := parseJsonRpcMessage(msg) |
| 427 | if msgId != nil { |
| 428 | pendingMu.Lock() |
| 429 | if req, exists := pendingRequests[msgId]; exists { |
| 430 | // Stop the timer |
| 431 | if req.Timer != nil { |
| 432 | req.Timer.Stop() |
| 433 | } |
| 434 | delete(pendingRequests, msgId) |
| 435 | } |
| 436 | pendingMu.Unlock() |
| 437 | } |
| 438 | } |
| 439 | } |
| 440 | messageQueue = nil // Clear the queue |
| 441 | } |
| 442 | messageQueueMu.Unlock() |
| 443 | |
| 444 | // Memory management: Limit size of pendingRequests to prevent memory leaks |
| 445 | pendingMu.Lock() |
| 446 | if len(pendingRequests) > 1000 { |
| 447 | fmt.Fprintf(os.Stderr, "%s: Too many pending requests (%d), cleaning up older ones\n", programName, len(pendingRequests)) |
| 448 | // Remove older entries (this is a bit tricky in Go without order guarantee) |
| 449 | // For simplicity, we'll just clear it all in this extreme case |
| 450 | pendingRequests = make(map[interface{}]*PendingRequest) |
| 451 | } |
| 452 | pendingMu.Unlock() |
| 453 | |
| 454 | // Signal for connection closure |
| 455 | connClosed := make(chan struct{}) |
| 456 | |
| 457 | // Create wait group for goroutines |
| 458 | var wg sync.WaitGroup |
| 459 | |
| 460 | // Goroutine for handling termination signals during connection |
| 461 | wg.Add(1) |
| 462 | go func() { |
| 463 | defer wg.Done() |
| 464 | select { |
| 465 | case <-ctx.Done(): |
| 466 | return |
| 467 | case <-doneCh: |
| 468 | // Clean up the connection |
| 469 | conn.Close(websocket.StatusNormalClosure, "Shutdown requested") |
| 470 | cancel() |
| 471 | return |
| 472 | case <-stdinClosedCh: |
| 473 | // Stdin closed, but keep connection active |
| 474 | stdinClosedCh = nil |
| 475 | return |
| 476 | } |
| 477 | }() |
| 478 | |
| 479 | // Goroutine for sending messages to websocket |
| 480 | wg.Add(1) |
| 481 | go func() { |
| 482 | defer wg.Done() |
| 483 | for { |
| 484 | select { |
| 485 | case <-ctx.Done(): |
| 486 | return |
| 487 | case <-connClosed: |
| 488 | return |
| 489 | case msg, ok := <-stdinCh: |
| 490 | if !ok { |
| 491 | // Stdin channel was closed |
| 492 | return |
| 493 | } |
| 494 | |
| 495 | err := conn.Write(ctx, websocket.MessageText, []byte(msg)) |
| 496 | if err != nil { |
| 497 | // Check if this was a message with ID |
| 498 | msgId, _ := parseJsonRpcMessage(msg) |
| 499 | |
| 500 | select { |
| 501 | case <-connClosed: |
| 502 | // Connection already known to be closed, queue the message |
| 503 | messageQueueMu.Lock() |
| 504 | messageQueue = append(messageQueue, msg) |
| 505 | messageQueueMu.Unlock() |
| 506 | |
| 507 | // If the message had an ID, keep the pending status |
| 508 | if msgId != nil { |
| 509 | pendingMu.Lock() |
| 510 | if _, exists := pendingRequests[msgId]; !exists { |
| 511 | // Create a new pending request with timeout if it doesn't exist |
| 512 | req := &PendingRequest{ |
| 513 | ID: msgId, |
| 514 | Message: msg, |
| 515 | Timer: time.AfterFunc(ConnectionTimeout, func() { handleRequestTimeout(msgId) }), |
| 516 | } |
| 517 | pendingRequests[msgId] = req |
| 518 | } |
| 519 | pendingMu.Unlock() |
| 520 | } |
| 521 | default: |
| 522 | fmt.Fprintf(os.Stderr, "%s: ERROR: write to websocket: %v\n", programName, err) |
| 523 | messageQueueMu.Lock() |
| 524 | messageQueue = append(messageQueue, msg) |
| 525 | messageQueueMu.Unlock() |
| 526 | |
| 527 | // Same ID handling as above |
| 528 | if msgId != nil { |
| 529 | pendingMu.Lock() |
| 530 | if _, exists := pendingRequests[msgId]; !exists { |
| 531 | req := &PendingRequest{ |
| 532 | ID: msgId, |
| 533 | Message: msg, |
| 534 | Timer: time.AfterFunc(ConnectionTimeout, func() { handleRequestTimeout(msgId) }), |
| 535 | } |
| 536 | pendingRequests[msgId] = req |
| 537 | } |
| 538 | pendingMu.Unlock() |
| 539 | } |
| 540 | } |
| 541 | } else { |
| 542 | // Message was sent successfully, remove from pending if it had an ID |
| 543 | msgId, _ := parseJsonRpcMessage(msg) |
| 544 | if msgId != nil { |
| 545 | pendingMu.Lock() |
| 546 | if req, exists := pendingRequests[msgId]; exists { |
| 547 | // Stop the connection establishment timer (if any) |
| 548 | if req.Timer != nil { |
| 549 | req.Timer.Stop() |
| 550 | req.Timer = nil // Clear timer to prevent memory leaks |
| 551 | } |
| 552 | // Keep in pendingRequests until we get a response |
| 553 | } |
| 554 | pendingMu.Unlock() |
| 555 | } |
| 556 | } |
| 557 | } |
| 558 | } |
| 559 | }() |
| 560 | |
| 561 | // Goroutine for receiving messages from websocket |
| 562 | wg.Add(1) |
| 563 | go func() { |
| 564 | defer wg.Done() |
| 565 | defer close(connClosed) |
| 566 | |
| 567 | for { |
| 568 | messageType, message, err := conn.Read(ctx) |
| 569 | if err != nil { |
| 570 | // Don't report error if context was cancelled |
| 571 | if ctx.Err() == nil { |
| 572 | fmt.Fprintf(os.Stderr, "%s: ERROR: read from websocket: %v\n", programName, err) |
| 573 | } |
| 574 | return |
| 575 | } |
| 576 | |
| 577 | if messageType == websocket.MessageText { |
| 578 | fmt.Println(string(message)) |
| 579 | |
| 580 | // Check if this is a response to a request with an ID |
| 581 | var response JsonRpcMessage |
| 582 | if err := json.Unmarshal(message, &response); err == nil && response.JsonRpc == "2.0" && response.Id != nil { |
| 583 | pendingMu.Lock() |
| 584 | if req, exists := pendingRequests[response.Id]; exists { |
| 585 | // Stop the timer |
| 586 | if req.Timer != nil { |
| 587 | req.Timer.Stop() |
| 588 | } |
| 589 | delete(pendingRequests, response.Id) |
| 590 | } |
| 591 | pendingMu.Unlock() |
| 592 | } |
| 593 | } |
| 594 | } |
| 595 | }() |
| 596 | |
| 597 | // Set up periodic pinger to keep connection alive |
| 598 | pingInterval := 30 * time.Second |
| 599 | wg.Add(1) |
| 600 | go func() { |
| 601 | defer wg.Done() |
| 602 | ticker := time.NewTicker(pingInterval) |
| 603 | defer ticker.Stop() |
| 604 | |
| 605 | for { |
| 606 | select { |
| 607 | case <-ticker.C: |
| 608 | // Send ping using a text message with a special ping format |
| 609 | err := conn.Write(ctx, websocket.MessageText, []byte("PING")) |
| 610 | if err != nil { |
| 611 | fmt.Fprintf(os.Stderr, "%s: ERROR: ping failed: %v\n", programName, err) |
| 612 | // Connection will be closed by the read loop |
| 613 | } |
| 614 | case <-ctx.Done(): |
| 615 | return |
| 616 | case <-connClosed: |
| 617 | return |
| 618 | } |
| 619 | } |
| 620 | }() |
| 621 | |
| 622 | // Wait for connection closed |
| 623 | <-connClosed |
| 624 | |
| 625 | // Update state to disconnected |
| 626 | stateMu.Lock() |
| 627 | state = stateDisconnected |
| 628 | stateMu.Unlock() |
| 629 | |
| 630 | // Clean up |
| 631 | cancel() |
| 632 | wg.Wait() |
| 633 | |
| 634 | // Don't automatically reconnect if stdin closed and no messages in queue |
| 635 | messageQueueMu.Lock() |
| 636 | hasMessages := len(messageQueue) > 0 |
| 637 | messageQueueMu.Unlock() |
| 638 | |
| 639 | if !stdinActive && !hasMessages { |
| 640 | fmt.Fprintf(os.Stderr, "%s: Stdin closed and no pending messages, exiting\n", programName) |
| 641 | return |
| 642 | } |
| 643 | |
| 644 | // Increment attempt counter for next reconnection |
| 645 | attempt++ |
| 646 | } |
| 647 | } |
| 648 | |
| 649 | // jitter returns a random value between 0.0 and 1.0 |
| 650 | func jitter() float64 { |
| 651 | n, err := rand.Int(rand.Reader, big.NewInt(1000)) |
| 652 | if err != nil { |
| 653 | return 0.5 // Fall back to 0.5 if there's an error |
| 654 | } |
| 655 | return float64(n.Int64()) / 1000.0 |
| 656 | } |