add keepAlive flag to prevent container from shutting down for long processes

.changeset/wet-falcons-hang.md

@@ -0,0 +1,5 @@
+---
+"@cloudflare/sandbox": patch
+---
+
+add keepAlive flag to prevent containers from shutting down

.github/workflows/pkg-pr-new.yml

@@ -8,6 +8,7 @@ on:
   pull_request:
     types: [opened, synchronize, reopened]
     paths:
+      - '**'
       - '!**/*.md'
       - '!.changeset/**'
 

packages/sandbox/src/request-handler.ts

@@ -1,5 +1,5 @@
-import { createLogger, type LogContext, TraceContext } from "@repo/shared";
 import { switchPort } from "@cloudflare/containers";
+import { createLogger, type LogContext, TraceContext } from "@repo/shared";
 import { getSandbox, type Sandbox } from "./sandbox";
 import {
   sanitizeSandboxId,

packages/sandbox/src/sandbox.ts

@@ -49,6 +49,10 @@ export function getSandbox(
     stub.setSleepAfter(options.sleepAfter);
   }
 
+  if (options?.keepAlive !== undefined) {
+    stub.setKeepAlive(options.keepAlive);
+  }
+
   return stub;
 }
 
@@ -64,6 +68,7 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
   private defaultSession: string | null = null;
   envVars: Record<string, string> = {};
   private logger: ReturnType<typeof createLogger>;
+  private keepAliveEnabled: boolean = false;
 
   constructor(ctx: DurableObject['ctx'], env: Env) {
     super(ctx, env);
@@ -131,6 +136,16 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
     this.sleepAfter = sleepAfter;
   }
 
+  // RPC method to enable keepAlive mode
+  async setKeepAlive(keepAlive: boolean): Promise<void> {
+    this.keepAliveEnabled = keepAlive;
+    if (keepAlive) {
+      this.logger.info('KeepAlive mode enabled - container will stay alive until explicitly destroyed');
+    } else {
+      this.logger.info('KeepAlive mode disabled - container will timeout normally');
+    }
+  }
+
   // RPC method to set environment variables
   async setEnvVars(envVars: Record<string, string>): Promise<void> {
     // Update local state for new sessions
@@ -220,6 +235,22 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
     this.logger.error('Sandbox error', error instanceof Error ? error : new Error(String(error)));
   }
 
+  /**
+   * Override onActivityExpired to prevent automatic shutdown when keepAlive is enabled
+   * When keepAlive is disabled, calls parent implementation which stops the container
+   */
+  override async onActivityExpired(): Promise<void> {
+    if (this.keepAliveEnabled) {
+      this.logger.debug('Activity expired but keepAlive is enabled - container will stay alive');
+      // Do nothing - don't call stop(), container stays alive
+    } else {
+      // Default behavior: stop the container
+      this.logger.debug('Activity expired - stopping container');
+      await super.onActivityExpired();
+    }
+  }
+
+
   // Override fetch to route internal container requests to appropriate ports
   override async fetch(request: Request): Promise<Response> {
     // Extract or generate trace ID from request
@@ -327,7 +358,6 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
     const startTime = Date.now();
     const timestamp = new Date().toISOString();
 
-    // Handle timeout
     let timeoutId: NodeJS.Timeout | undefined;
 
     try {
@@ -592,8 +622,7 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
     };
   }
 
-
-  // Streaming methods - return ReadableStream for RPC compatibility
+// Streaming methods - return ReadableStream for RPC compatibility
   async execStream(command: string, options?: StreamOptions): Promise<ReadableStream<Uint8Array>> {
     // Check for cancellation
     if (options?.signal?.aborted) {
@@ -617,6 +646,9 @@ export class Sandbox<Env = unknown> extends Container<Env> implements ISandbox {
     return this.client.commands.executeStream(command, sessionId);
   }
 
+  /**
+   * Stream logs from a background process as a ReadableStream.
+   */
   async streamProcessLogs(processId: string, options?: { signal?: AbortSignal }): Promise<ReadableStream<Uint8Array>> {
     // Check for cancellation
     if (options?.signal?.aborted) {

packages/sandbox/tests/get-sandbox.test.ts

@@ -30,6 +30,7 @@ describe('getSandbox', () => {
       setSleepAfter: vi.fn((value: string | number) => {
         mockStub.sleepAfter = value;
       }),
+      setKeepAlive: vi.fn(),
     };
 
     // Mock getContainer to return our stub
@@ -107,4 +108,42 @@ describe('getSandbox', () => {
       expect(sandbox.sleepAfter).toBe(timeString);
     }
   });
+
+  it('should apply keepAlive option when provided as true', () => {
+    const mockNamespace = {} as any;
+    const sandbox = getSandbox(mockNamespace, 'test-sandbox', {
+      keepAlive: true,
+    });
+
+    expect(sandbox.setKeepAlive).toHaveBeenCalledWith(true);
+  });
+
+  it('should apply keepAlive option when provided as false', () => {
+    const mockNamespace = {} as any;
+    const sandbox = getSandbox(mockNamespace, 'test-sandbox', {
+      keepAlive: false,
+    });
+
+    expect(sandbox.setKeepAlive).toHaveBeenCalledWith(false);
+  });
+
+  it('should not call setKeepAlive when keepAlive option not provided', () => {
+    const mockNamespace = {} as any;
+    getSandbox(mockNamespace, 'test-sandbox');
+
+    expect(mockStub.setKeepAlive).not.toHaveBeenCalled();
+  });
+
+  it('should apply keepAlive alongside other options', () => {
+    const mockNamespace = {} as any;
+    const sandbox = getSandbox(mockNamespace, 'test-sandbox', {
+      sleepAfter: '5m',
+      baseUrl: 'https://example.com',
+      keepAlive: true,
+    });
+
+    expect(sandbox.sleepAfter).toBe('5m');
+    expect(sandbox.setBaseUrl).toHaveBeenCalledWith('https://example.com');
+    expect(sandbox.setKeepAlive).toHaveBeenCalledWith(true);
+  });
 });

packages/sandbox/tests/sandbox.test.ts

@@ -1,7 +1,7 @@
+import { Container } from '@cloudflare/containers';
 import type { DurableObjectState } from '@cloudflare/workers-types';
 import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest';
 import { Sandbox } from '../src/sandbox';
-import { Container } from '@cloudflare/containers';
 
 // Mock dependencies before imports
 vi.mock('./interpreter', () => ({
@@ -48,7 +48,7 @@ describe('Sandbox - Automatic Session Management', () => {
         delete: vi.fn().mockResolvedValue(undefined),
         list: vi.fn().mockResolvedValue(new Map()),
       } as any,
-      blockConcurrencyWhile: vi.fn((fn: () => Promise<void>) => fn()),
+      blockConcurrencyWhile: vi.fn().mockImplementation(<T>(callback: () => Promise<T>): Promise<T> => callback()),
       id: {
         toString: () => 'test-sandbox-id',
         equals: vi.fn(),

packages/shared/src/index.ts

@@ -54,9 +54,9 @@ export type {
   ExecOptions,
   ExecResult,
   ExecutionSession,
-  FileExistsResult,
   // File streaming types
   FileChunk,
+  FileExistsResult,
   FileInfo,
   FileMetadata,
   FileStreamEvent,

packages/shared/src/types.ts

@@ -262,13 +262,26 @@ export interface SandboxOptions {
    * - A string like "30s", "3m", "5m", "1h" (seconds, minutes, or hours)
    * - A number representing seconds (e.g., 180 for 3 minutes)
    * Default: "10m" (10 minutes)
+   *
+   * Note: Ignored when keepAlive is true
    */
   sleepAfter?: string | number;
 
   /**
    * Base URL for the sandbox API
    */
   baseUrl?: string;
+
+  /**
+   * Keep the container alive indefinitely by preventing automatic shutdown
+   * When true, the container will never auto-timeout and must be explicitly destroyed
+   * - Any scenario where activity can't be automatically detected
+   *
+   * Important: You MUST call sandbox.destroy() when done to avoid resource leaks
+   *
+   * Default: false
+   */
+  keepAlive?: boolean;
 }
 
 /**
@@ -590,7 +603,7 @@ export interface ExecutionSession {
   // Command execution
   exec(command: string, options?: ExecOptions): Promise<ExecResult>;
   execStream(command: string, options?: StreamOptions): Promise<ReadableStream<Uint8Array>>;
-  
+
   // Background process management
   startProcess(command: string, options?: ProcessOptions): Promise<Process>;
   listProcesses(): Promise<Process[]>;
@@ -621,7 +634,7 @@ export interface ExecutionSession {
   // Code interpreter methods
   createCodeContext(options?: CreateContextOptions): Promise<CodeContext>;
   runCode(code: string, options?: RunCodeOptions): Promise<ExecutionResult>;
-  runCodeStream(code: string, options?: RunCodeOptions): Promise<ReadableStream>;
+  runCodeStream(code: string, options?: RunCodeOptions): Promise<ReadableStream<Uint8Array>>;
   listCodeContexts(): Promise<CodeContext[]>;
   deleteCodeContext(contextId: string): Promise<void>;
 }