What is the Retry Pattern?

The Retry Pattern automatically re-attempts a failed operation, assuming the failure is transient (temporary). Network blips, timeouts, and brief service outages often resolve themselves within seconds.

When to Retry

Retry	Don't Retry
Network timeout	400 Bad Request
503 Service Unavailable	401 Unauthorized
429 Too Many Requests	404 Not Found
Connection reset	Business logic error
DNS resolution failure	Data validation error

Retry Strategies

Immediate Retry

Immediate RetryFlowchart

Simple but can overwhelm an already struggling service.

Fixed Delay

Attempt 1 → fail → wait 2s
Attempt 2 → fail → wait 2s
Attempt 3 → fail → wait 2s
Attempt 4 → fail → give up

Exponential Backoff

Attempt 1 → fail → wait 1s
Attempt 2 → fail → wait 2s
Attempt 3 → fail → wait 4s
Attempt 4 → fail → wait 8s
Attempt 5 → fail → give up

Exponential Backoff with Jitter

Attempt 1 → fail → wait 1s + random(0-500ms)
Attempt 2 → fail → wait 2s + random(0-500ms)
Attempt 3 → fail → wait 4s + random(0-500ms)

Jitter prevents the thundering herd problem where all clients retry simultaneously.

Comparison

Strategy	Delay	Best For
Immediate	None	Very transient failures
Fixed	Constant	Known recovery times
Exponential	Increasing	Unknown recovery times
Exponential + Jitter	Increasing + random	High-concurrency systems

The Thundering Herd Problem

The Thundering Herd ProblemSequence diagram

Code Example

async function retryWithBackoff(fn, options = {}) {
  const {
    maxRetries = 3,
    baseDelay = 1000,
    maxDelay = 30000,
    jitter = true
  } = options;

  for (let attempt = 0; attempt <= maxRetries; attempt++) {
    try {
      return await fn();
    } catch (error) {
      if (attempt === maxRetries || !isRetryable(error)) {
        throw error;
      }

      let delay = Math.min(baseDelay * Math.pow(2, attempt), maxDelay);
      if (jitter) {
        delay += Math.random() * delay * 0.5;
      }

      await new Promise(resolve => setTimeout(resolve, delay));
    }
  }
}

function isRetryable(error) {
  const retryableCodes = [408, 429, 500, 502, 503, 504];
  return retryableCodes.includes(error.statusCode);
}

Retry + Circuit Breaker

These patterns complement each other:

Retry + Circuit BreakerFlowchart

Retry Budgets

Limit total retries across all clients:

Approach	Description
Per-request limit	Max 3 retries per request
Per-client limit	Max 10 retries per second
System-wide budget	Total retries < 10% of requests

Best Practices

Only retry idempotent operations (or use idempotency keys)
Set a maximum retry count to avoid infinite loops
Use exponential backoff with jitter for distributed systems
Log retry attempts for debugging
Combine with circuit breaker to fail fast when service is down

Interview Tips

Explain exponential backoff with jitter
Discuss which errors are retryable vs permanent
Mention thundering herd problem
Combine with circuit breaker pattern
Know retry budgets for system-wide protection

Finished reading?

Mark it done to track your progress in Concepts.

Retry Pattern

On this page