How to Implement Retry with Exponential Backoff for Cloud API 429 Errors in Python

Implement a retry-with-backoff loop in Python to handle 429 throttling errors from cloud APIs, using exponential delay between attempts.

Easy Python 3.9+ Aug 9, 2026 Cloud + Python 13 views 0 copies

retry backoff 429 rate-limit requests

Requires third-party packages — install first

pip install requests

Python code

30 lines

Python 3.9+

import time
import random
import requests


def api_call(attempt):
    """Mock cloud API that returns 429 for the first two attempts."""
    if attempt < 2:
        return 429, "Too Many Requests"
    return 200, {"data": "success"}


def retry_with_backoff(api_func, max_retries=3, base_delay=0.1):
    for attempt in range(max_retries):
        status, body = api_func(attempt)
        if status == 429:
            delay = base_delay * (2 ** attempt)
            print(f"Attempt {attempt + 1}: 429, retrying in {delay:.2f}s")
            time.sleep(delay)
        else:
            print(f"Attempt {attempt + 1}: {status} - {body}")
            return status, body
    print("Max retries exceeded")
    return None


if __name__ == "__main__":
    random.seed(42)
    result = retry_with_backoff(lambda attempt: api_call(attempt))
    print("Final result:", result)

Output

stdout

Attempt 1: 429, retrying in 0.10s
Attempt 2: 429, retrying in 0.20s
Attempt 3: 200 - {'data': 'success'}
Final result: (200, {'data': 'success'})

How it works

The retry_with_backoff function loops through a maximum number of attempts, calling the API function each time. When the status code is 429, it calculates a delay using exponential backoff (base_delay * (2 ** attempt)) and sleeps to respect the throttle. If the call succeeds (status != 429), it returns the result immediately, avoiding unnecessary waits. The loop tracks the attempt index so the delay grows with each retry, a standard pattern in cloud SDKs. Using random.seed(42) here is optional—it just makes the example reproducible, but the actual retry logic does not rely on randomness.

Common mistakes

Not sleeping between retries, causing immediate hammering of the API
Using a fixed delay instead of exponential backoff, which increases load on already-throttled services
Forgetting to handle other error status codes (e.g., 5xx) that should also be retried
Exceeding resource limits by retrying indefinitely without a max attempts cap

Variations

Use `tenacity` library with `@retry(wait=wait_exponential())` to simplify retry logic
Parse the `Retry-After` header from the 429 response to set a custom delay

Real-world use cases

Making authenticated calls to AWS, GCP, or Azure APIs that rate-limit requests per second
Fetching data from REST endpoints like GitHub or Stripe that return 429 when quotas are exceeded
Processing large batches of API requests in a data pipeline where throttling is common

How to Implement Retry with Exponential Backoff for Cloud API 429 Errors in Python

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Cloud + Python

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Cloud + Python

Keep learning

Tutorials

Quizzes