Redux Saga报错后停止运行，如何在RootSaga全局捕获错误？

Great question! The core issue here is how Redux Saga's yield all behaves when one of the parallel sagas throws an uncaught error—by default, it cancels all other sagas and terminates the root saga entirely. To fix this, we need to isolate each module's saga with a global error handler so that individual failures don't take down the whole system.

Step 1: Create a Global Error Handling Wrapper #

First, define a utility function that wraps any saga with a try/catch block. This will capture errors for the individual saga and let you add global error logic (like logging or alerts) without affecting other sagas:

// Global error handler wrapper for sagas
const wrapSagaWithGlobalErrorHandler = (saga) => function* wrappedSaga() {
  try {
    // Execute the original saga
    yield call(saga);
  } catch (error) {
    // Add your custom global error handling here
    console.error('[Global Saga Error] Uncaught error in saga:', error);
    console.error('Error stack trace:', error.stack);
    
    // Optional: Auto-restart the failed saga (uncomment if needed)
    // console.log('[Global Saga Restart] Restarting failed module saga...');
    // yield call(wrapSagaWithGlobalErrorHandler, saga);
  }
};

Step 2: Modify Your Root Saga to Use the Wrapper #

Update your root saga to wrap every module saga with the error handler before passing them to yield all:

export default function* rootSaga() {
  // Wrap each module saga with the global error handler
  const safeSagas = sagas.map(saga => wrapSagaWithGlobalErrorHandler(saga));
  
  // Run all wrapped sagas in parallel
  yield all(safeSagas);
}

How This Works #

• Isolation: Each module's saga runs in its own try/catch context. If one throws an error, it's caught locally instead of propagating up to yield all.
• Non-disruptive: Other sagas continue running normally even if one fails. Your root saga stays active.
• Flexibility: You can extend the catch block to add error reporting (like sending logs to a monitoring service) or user-facing alerts. The optional auto-restart logic ensures a failed saga can recover automatically.

Notes for Edge Cases #

• For takeEvery/takeLatest workers: By default, errors in these workers are isolated to the forked task and won't crash the parent saga. However, if your module saga has initialization errors (e.g., calling an undefined function before setting up listeners), the wrapper will catch those.
• If you need per-saga error customization: You can modify the wrapper to accept an optional error handler, but the global wrapper covers the "catch-all" use case for unhandled errors across your multi-team project.

内容的提问来源于stack exchange，提问作者Milindu Sanoj Kumarage