How to Compose Parallel API Calls in Python with asyncio.gather

Compose multiple mock API responses in parallel using asyncio.gather with per-service simulated latency.

Easy Python 3.9+ Aug 9, 2026 Microservices patterns 15 views 0 copies

asyncio concurrency api microservices

Python code

23 lines

Python 3.9+

import asyncio
import random
import time

async def mock_api(name: str, delay: float) -> dict:
    await asyncio.sleep(delay)
    return {"service": name, "value": random.randint(1, 100)}

async def fetch_all():
    services = {
        "users": mock_api("users", 0.2),
        "orders": mock_api("orders", 0.3),
        "products": mock_api("products", 0.1),
    }
    start = time.perf_counter()
    results = await asyncio.gather(*services.values(), return_exceptions=True)
    elapsed = time.perf_counter() - start
    combined = {name: result for name, result in zip(services.keys(), results)}
    print(f"Total time: {elapsed:.2f}s")
    print(f"Composed result: {combined}")

if __name__ == "__main__":
    asyncio.run(fetch_all())

Output

stdout

Total time: 0.31s
Composed result: {'users': {'service': 'users', 'value': 42}, 'orders': {'service': 'orders', 'value': 87}, 'products': {'service': 'products', 'value': 15}}

How it works

The code creates three coroutines for mock services and schedules them all at once inside a dict. asyncio.gather runs them concurrently, so the wall-clock time equals the slowest request (~0.3s) rather than the sum of all delays. The return_exceptions=True flag prevents one failing service from cancelling the others. Because gather preserves input order, zipping the service keys with the results maps each response to its correct origin.

Common mistakes

Calling the async function immediately in the dict, which schedules it; use `functools.partial` or lambdas if arguments differ per call.
Forgetting `return_exceptions=True`, which makes a single API timeout crash the whole composition.
Assuming `gather` returns results in completion order instead of input order.

Variations

Use `asyncio.create_task` plus `asyncio.wait` for fine-grained control over timeouts and cancellations.
Switch to `httpx.AsyncClient` with `.get` calls instead of mocks for real HTTP requests.

Real-world use cases

A backend BFF (Backend-for-Frontend) aggregating user profiles, orders, and product data into one response.
A payment orchestration service fetching customer, balance, and transaction history in parallel before decisioning.
A metrics dashboard pulling health status from multiple microservices in a single refresh call.

How to Compose Parallel API Calls in Python with asyncio.gather

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Microservices patterns

Tutorials

Quizzes

Python code

Output

How it works

Common mistakes

Variations

Real-world use cases

More from Microservices patterns

Keep learning

Tutorials

Quizzes