feat: Refactor CaddyManager and OneboxReverseProxy to use Docker service for Caddy management

ts/classes/caddy.ts

@@ -1,25 +1,27 @@
 /**
  * Caddy Manager for Onebox
  *
- * Manages Caddy binary download, process lifecycle, and Admin API configuration.
- * Caddy is used as the reverse proxy with native SNI support.
+ * Manages Caddy as a Docker Swarm service instead of a host binary.
+ * This allows Caddy to access services on the Docker overlay network.
  */
 
+import * as plugins from '../plugins.ts';
 import { logger } from '../logging.ts';
 import { getErrorMessage } from '../utils/error.ts';
 
-const CADDY_VERSION = '2.10.2';
-const CADDY_DOWNLOAD_URL = `https://github.com/caddyserver/caddy/releases/download/v${CADDY_VERSION}/caddy_${CADDY_VERSION}_linux_amd64.tar.gz`;
+const CADDY_SERVICE_NAME = 'onebox-caddy';
+const CADDY_IMAGE = 'caddy:2-alpine';
+const DOCKER_GATEWAY_IP = '172.17.0.1'; // Docker bridge gateway for container-to-host communication
 
 export interface ICaddyRoute {
   domain: string;
-  upstream: string; // e.g., "10.0.1.40:80"
+  upstream: string; // e.g., "onebox-hello-world:80"
 }
 
 export interface ICaddyCertificate {
   domain: string;
-  certPath: string;
-  keyPath: string;
+  certPem: string;
+  keyPem: string;
 }
 
 interface ICaddyLoggingConfig {
@@ -64,7 +66,7 @@ interface ICaddyConfig {
         policies: Array<{ issuers: never[] }>;
       };
       certificates?: {
-        load_files?: Array<{
+        load_pem?: Array<{
           certificate: string;
           key: string;
           tags?: string[];
@@ -85,8 +87,7 @@ interface ICaddyRouteConfig {
 }
 
 export class CaddyManager {
-  private process: Deno.ChildProcess | null = null;
-  private binaryPath: string;
+  private dockerClient: InstanceType<typeof plugins.docker.Docker> | null = null;
   private certsDir: string;
   private adminUrl: string;
   private httpPort: number;
@@ -95,9 +96,10 @@ export class CaddyManager {
   private loggingEnabled: boolean;
   private routes: Map<string, ICaddyRoute> = new Map();
   private certificates: Map<string, ICaddyCertificate> = new Map();
+  private networkName = 'onebox-network';
+  private serviceRunning = false;
 
   constructor(options?: {
-    binaryPath?: string;
     certsDir?: string;
     adminPort?: number;
     httpPort?: number;
@@ -105,7 +107,6 @@ export class CaddyManager {
     logReceiverPort?: number;
     loggingEnabled?: boolean;
   }) {
-    this.binaryPath = options?.binaryPath || './.nogit/caddy';
     this.certsDir = options?.certsDir || './.nogit/certs';
     this.adminUrl = `http://localhost:${options?.adminPort || 2019}`;
     this.httpPort = options?.httpPort || 8080;
@@ -114,6 +115,18 @@ export class CaddyManager {
     this.loggingEnabled = options?.loggingEnabled ?? true;
   }
 
+  /**
+   * Initialize Docker client for Caddy service management
+   */
+  private async ensureDockerClient(): Promise<void> {
+    if (!this.dockerClient) {
+      this.dockerClient = new plugins.docker.Docker({
+        socketPath: 'unix:///var/run/docker.sock',
+      });
+      await this.dockerClient.start();
+    }
+  }
+
   /**
    * Update listening ports (must call reloadConfig after if running)
    */
@@ -123,103 +136,98 @@ export class CaddyManager {
   }
 
   /**
-   * Ensure Caddy binary is downloaded and executable
-   */
-  async ensureBinary(): Promise<void> {
-    try {
-      // Check if binary exists
-      try {
-        const stat = await Deno.stat(this.binaryPath);
-        if (stat.isFile) {
-          // Verify it's executable by checking version
-          const cmd = new Deno.Command(this.binaryPath, {
-            args: ['version'],
-            stdout: 'piped',
-            stderr: 'piped',
-          });
-          const result = await cmd.output();
-          if (result.success) {
-            const version = new TextDecoder().decode(result.stdout).trim();
-            logger.info(`Caddy binary found: ${version}`);
-            return;
-          }
-        }
-      } catch {
-        // Binary doesn't exist, need to download
-      }
-
-      logger.info(`Downloading Caddy v${CADDY_VERSION}...`);
-
-      // Create directory if needed
-      const dir = this.binaryPath.substring(0, this.binaryPath.lastIndexOf('/'));
-      await Deno.mkdir(dir, { recursive: true });
-
-      // Download tar.gz
-      const response = await fetch(CADDY_DOWNLOAD_URL);
-      if (!response.ok) {
-        throw new Error(`Failed to download Caddy: ${response.status} ${response.statusText}`);
-      }
-
-      const tarGzPath = `${this.binaryPath}.tar.gz`;
-      const data = new Uint8Array(await response.arrayBuffer());
-      await Deno.writeFile(tarGzPath, data);
-
-      // Extract using tar command
-      const extractCmd = new Deno.Command('tar', {
-        args: ['-xzf', tarGzPath, '-C', dir, 'caddy'],
-        stdout: 'piped',
-        stderr: 'piped',
-      });
-      const extractResult = await extractCmd.output();
-      if (!extractResult.success) {
-        throw new Error(`Failed to extract Caddy: ${new TextDecoder().decode(extractResult.stderr)}`);
-      }
-
-      // Clean up tar.gz
-      await Deno.remove(tarGzPath);
-
-      // Make executable
-      await Deno.chmod(this.binaryPath, 0o755);
-
-      logger.success(`Caddy v${CADDY_VERSION} downloaded to ${this.binaryPath}`);
-    } catch (error) {
-      logger.error(`Failed to ensure Caddy binary: ${getErrorMessage(error)}`);
-      throw error;
-    }
-  }
-
-  /**
-   * Start Caddy process with minimal config, then configure via Admin API
+   * Start Caddy as a Docker Swarm service
    */
   async start(): Promise<void> {
-    if (this.process) {
-      logger.warn('Caddy is already running');
+    if (this.serviceRunning) {
+      logger.warn('Caddy service is already running');
       return;
     }
 
     try {
-      // Create certs directory
+      await this.ensureDockerClient();
+
+      // Create certs directory for backup/persistence
       await Deno.mkdir(this.certsDir, { recursive: true });
 
-      logger.info('Starting Caddy server...');
+      logger.info('Starting Caddy Docker service...');
 
-      // Start Caddy with blank config - Admin API is available immediately
-      // We'll push the full configuration via Admin API after it's ready
-      const cmd = new Deno.Command(this.binaryPath, {
-        args: ['run'],
-        stdin: 'null',
-        stdout: 'piped',
-        stderr: 'piped',
+      // Check if service already exists
+      const existingService = await this.getExistingService();
+      if (existingService) {
+        logger.info('Caddy service exists, removing old service...');
+        await this.removeService();
+        // Wait for service to be removed
+        await new Promise((resolve) => setTimeout(resolve, 2000));
+      }
+
+      // Get network ID
+      const networkId = await this.getNetworkId();
+
+      // Create Caddy Docker service
+      const response = await this.dockerClient!.request('POST', '/services/create', {
+        Name: CADDY_SERVICE_NAME,
+        Labels: {
+          'managed-by': 'onebox',
+          'onebox-type': 'caddy',
+        },
+        TaskTemplate: {
+          ContainerSpec: {
+            Image: CADDY_IMAGE,
+            // Start Caddy with admin listening on all interfaces so we can reach it from host
+            // Write minimal config to /tmp and start Caddy with that config
+            Command: ['sh', '-c', 'printf \'{"admin":{"listen":"0.0.0.0:2019"}}\' > /tmp/caddy.json && caddy run --config /tmp/caddy.json'],
+          },
+          Networks: [
+            {
+              Target: networkId,
+            },
+          ],
+          RestartPolicy: {
+            Condition: 'any',
+            MaxAttempts: 0,
+          },
+        },
+        Mode: {
+          Replicated: {
+            Replicas: 1,
+          },
+        },
+        EndpointSpec: {
+          Ports: [
+            {
+              Protocol: 'tcp',
+              TargetPort: 80,
+              PublishedPort: this.httpPort,
+              PublishMode: 'host',
+            },
+            {
+              Protocol: 'tcp',
+              TargetPort: 443,
+              PublishedPort: this.httpsPort,
+              PublishMode: 'host',
+            },
+            {
+              Protocol: 'tcp',
+              TargetPort: 2019,
+              PublishedPort: 2019,
+              PublishMode: 'host',
+            },
+          ],
+        },
       });
 
-      this.process = cmd.spawn();
+      if (response.statusCode >= 300) {
+        throw new Error(`Failed to create Caddy service: HTTP ${response.statusCode} - ${JSON.stringify(response.body)}`);
+      }
 
-      // Start log readers (non-blocking)
-      this.readProcessOutput();
+      logger.info(`Caddy service created: ${response.body.ID}`);
 
       // Wait for Admin API to be ready
       await this.waitForReady();
 
+      this.serviceRunning = true;
+
       // Now configure via Admin API with current routes and certificates
       await this.reloadConfig();
 
@@ -231,42 +239,47 @@ export class CaddyManager {
   }
 
   /**
-   * Read process stdout/stderr and log
+   * Get existing Caddy service if any
    */
-  private async readProcessOutput(): Promise<void> {
-    if (!this.process) return;
-
-    // Read stderr (Caddy logs to stderr by default)
-    const stderrReader = this.process.stderr.getReader();
-    (async () => {
-      try {
-        while (true) {
-          const { done, value } = await stderrReader.read();
-          if (done) break;
-          const text = new TextDecoder().decode(value).trim();
-          if (text) {
-            // Parse Caddy's JSON log format or just log as-is
-            for (const line of text.split('\n')) {
-              if (line.includes('"level":"error"')) {
-                logger.error(`[Caddy] ${line}`);
-              } else if (line.includes('"level":"warn"')) {
-                logger.warn(`[Caddy] ${line}`);
-              } else {
-                logger.debug(`[Caddy] ${line}`);
-              }
-            }
-          }
-        }
-      } catch {
-        // Process ended
+  private async getExistingService(): Promise<any | null> {
+    try {
+      const response = await this.dockerClient!.request('GET', `/services/${CADDY_SERVICE_NAME}`, {});
+      if (response.statusCode === 200) {
+        return response.body;
       }
-    })();
+      return null;
+    } catch {
+      return null;
+    }
   }
 
   /**
-   * Wait for Caddy to be ready by polling admin API
+   * Remove the Caddy service
    */
-  private async waitForReady(maxAttempts = 50, intervalMs = 100): Promise<void> {
+  private async removeService(): Promise<void> {
+    try {
+      await this.dockerClient!.request('DELETE', `/services/${CADDY_SERVICE_NAME}`, {});
+    } catch {
+      // Service may not exist
+    }
+  }
+
+  /**
+   * Get network ID by name
+   */
+  private async getNetworkId(): Promise<string> {
+    const networks = await this.dockerClient!.listNetworks();
+    const network = networks.find((n: any) => n.Name === this.networkName);
+    if (!network) {
+      throw new Error(`Network not found: ${this.networkName}`);
+    }
+    return network.Id;
+  }
+
+  /**
+   * Wait for Caddy Admin API to be ready
+   */
+  private async waitForReady(maxAttempts = 60, intervalMs = 500): Promise<void> {
     for (let i = 0; i < maxAttempts; i++) {
       try {
         const response = await fetch(`${this.adminUrl}/config/`);
@@ -278,48 +291,33 @@ export class CaddyManager {
       }
       await new Promise((resolve) => setTimeout(resolve, intervalMs));
     }
-    throw new Error('Caddy failed to start within timeout');
+    throw new Error('Caddy service failed to start within timeout');
   }
 
   /**
-   * Stop Caddy process
+   * Stop Caddy Docker service
    */
   async stop(): Promise<void> {
-    if (!this.process) {
+    if (!this.serviceRunning && !(await this.getExistingService())) {
       return;
     }
 
     try {
-      logger.info('Stopping Caddy...');
+      await this.ensureDockerClient();
 
-      // Try graceful shutdown via API first
-      try {
-        await fetch(`${this.adminUrl}/stop`, { method: 'POST' });
-        // Wait for process to exit
-        await Promise.race([
-          this.process.status,
-          new Promise((resolve) => setTimeout(resolve, 5000)),
-        ]);
-      } catch {
-        // API not available, kill directly
-      }
+      logger.info('Stopping Caddy service...');
 
-      // Force kill if still running
-      try {
-        this.process.kill('SIGTERM');
-      } catch {
-        // Already dead
-      }
+      await this.removeService();
 
-      this.process = null;
-      logger.info('Caddy stopped');
+      this.serviceRunning = false;
+      logger.info('Caddy service stopped');
     } catch (error) {
       logger.error(`Failed to stop Caddy: ${getErrorMessage(error)}`);
     }
   }
 
   /**
-   * Check if Caddy is healthy
+   * Check if Caddy Admin API is healthy
    */
   async isHealthy(): Promise<boolean> {
     try {
@@ -330,6 +328,31 @@ export class CaddyManager {
     }
   }
 
+  /**
+   * Check if Caddy service is running
+   */
+  async isRunning(): Promise<boolean> {
+    try {
+      await this.ensureDockerClient();
+      const service = await this.getExistingService();
+      if (!service) return false;
+
+      // Check if service has running tasks
+      const tasksResponse = await this.dockerClient!.request(
+        'GET',
+        `/tasks?filters=${encodeURIComponent(JSON.stringify({ service: [CADDY_SERVICE_NAME] }))}`,
+        {}
+      );
+
+      if (tasksResponse.statusCode !== 200) return false;
+
+      const tasks = tasksResponse.body;
+      return tasks.some((task: any) => task.Status?.State === 'running');
+    } catch {
+      return false;
+    }
+  }
+
   /**
    * Build Caddy JSON configuration from current routes and certificates
    */
@@ -350,27 +373,27 @@ export class CaddyManager {
       });
     }
 
-    // Build certificate load_files
-    const loadFiles: Array<{ certificate: string; key: string; tags?: string[] }> = [];
+    // Build certificate load_pem entries (inline PEM content)
+    const loadPem: Array<{ certificate: string; key: string; tags?: string[] }> = [];
     for (const [domain, cert] of this.certificates) {
-      loadFiles.push({
-        certificate: cert.certPath,
-        key: cert.keyPath,
+      loadPem.push({
+        certificate: cert.certPem,
+        key: cert.keyPem,
         tags: [domain],
       });
     }
 
     const config: ICaddyConfig = {
       admin: {
-        listen: this.adminUrl.replace('http://', ''),
+        listen: '0.0.0.0:2019', // Listen on all interfaces inside container
       },
       apps: {
         http: {
           servers: {
             main: {
-              listen: [`:${this.httpPort}`, `:${this.httpsPort}`],
+              listen: [':80', ':443'],
               routes,
-              // Disable automatic HTTPS to prevent Caddy from trying to bind to port 80/443
+              // Disable automatic HTTPS to prevent Caddy from trying to obtain certs
               automatic_https: {
                 disable: true,
               },
@@ -387,7 +410,8 @@ export class CaddyManager {
           access: {
             writer: {
               output: 'net',
-              address: `tcp/localhost:${this.logReceiverPort}`,
+              // Use Docker bridge gateway IP to reach log receiver on host
+              address: `tcp/${DOCKER_GATEWAY_IP}:${this.logReceiverPort}`,
               dial_timeout: '5s',
               soft_start: true, // Continue even if log receiver is down
             },
@@ -405,14 +429,14 @@ export class CaddyManager {
     }
 
     // Add TLS config if we have certificates
-    if (loadFiles.length > 0) {
+    if (loadPem.length > 0) {
       config.apps.tls = {
         automation: {
           // Disable automatic HTTPS - we manage certs ourselves
           policies: [{ issuers: [] }],
         },
         certificates: {
-          load_files: loadFiles,
+          load_pem: loadPem,
         },
       };
     }
@@ -424,7 +448,8 @@ export class CaddyManager {
    * Reload Caddy configuration via Admin API
    */
   async reloadConfig(): Promise<void> {
-    if (!this.process) {
+    const isRunning = await this.isRunning();
+    if (!isRunning) {
       logger.warn('Caddy not running, cannot reload config');
       return;
     }
@@ -456,7 +481,7 @@ export class CaddyManager {
   async addRoute(domain: string, upstream: string): Promise<void> {
     this.routes.set(domain, { domain, upstream });
 
-    if (this.process) {
+    if (await this.isRunning()) {
       await this.reloadConfig();
     }
 
@@ -468,7 +493,7 @@ export class CaddyManager {
    */
   async removeRoute(domain: string): Promise<void> {
     if (this.routes.delete(domain)) {
-      if (this.process) {
+      if (await this.isRunning()) {
         await this.reloadConfig();
       }
       logger.success(`Removed Caddy route: ${domain}`);
@@ -477,28 +502,26 @@ export class CaddyManager {
 
   /**
    * Add or update a TLS certificate
-   * Writes PEM files to disk and updates config
+   * Stores PEM content in memory for Admin API, also writes to disk for backup
    */
   async addCertificate(domain: string, certPem: string, keyPem: string): Promise<void> {
-    // Write PEM files
-    const certPath = `${this.certsDir}/${domain}.crt`;
-    const keyPath = `${this.certsDir}/${domain}.key`;
-
-    await Deno.mkdir(this.certsDir, { recursive: true });
-    await Deno.writeTextFile(certPath, certPem);
-    await Deno.writeTextFile(keyPath, keyPem);
-
-    // Use absolute paths for Caddy
-    const absoluteCertPath = await Deno.realPath(certPath);
-    const absoluteKeyPath = await Deno.realPath(keyPath);
-
+    // Store PEM content in memory for buildConfig()
     this.certificates.set(domain, {
       domain,
-      certPath: absoluteCertPath,
-      keyPath: absoluteKeyPath,
+      certPem,
+      keyPem,
     });
 
-    if (this.process) {
+    // Also write to disk for backup/persistence
+    try {
+      await Deno.mkdir(this.certsDir, { recursive: true });
+      await Deno.writeTextFile(`${this.certsDir}/${domain}.crt`, certPem);
+      await Deno.writeTextFile(`${this.certsDir}/${domain}.key`, keyPem);
+    } catch (error) {
+      logger.warn(`Failed to write certificate backup for ${domain}: ${getErrorMessage(error)}`);
+    }
+
+    if (await this.isRunning()) {
       await this.reloadConfig();
     }
 
@@ -509,19 +532,16 @@ export class CaddyManager {
    * Remove a TLS certificate
    */
   async removeCertificate(domain: string): Promise<void> {
-    const cert = this.certificates.get(domain);
-    if (cert) {
-      this.certificates.delete(domain);
-
-      // Remove files
+    if (this.certificates.delete(domain)) {
+      // Remove backup files
       try {
-        await Deno.remove(cert.certPath);
-        await Deno.remove(cert.keyPath);
+        await Deno.remove(`${this.certsDir}/${domain}.crt`);
+        await Deno.remove(`${this.certsDir}/${domain}.key`);
       } catch {
         // Files may not exist
       }
 
-      if (this.process) {
+      if (await this.isRunning()) {
         await this.reloadConfig();
       }
 
@@ -562,7 +582,7 @@ export class CaddyManager {
     certificates: number;
   } {
     return {
-      running: this.process !== null,
+      running: this.serviceRunning,
       httpPort: this.httpPort,
       httpsPort: this.httpsPort,
       routes: this.routes.size,