Timeouts and Deadlines (Prevent Hanging)

Use timeouts and end-to-end deadlines to prevent stuck calls, bound tail latency, and keep systems responsive under partial failure.

On this page

Timeout vs Deadline

timeout: limit for one operation
deadline: total time budget across multiple steps

Budget the Whole Request

Pick an overall deadline and allocate sub-timeouts so downstream calls cannot consume the entire budget.

asyncio Timeout Example

import asyncio

async def call():
    await asyncio.sleep(10)

async def main():
    return await asyncio.wait_for(call(), timeout=0.5)

try:
    asyncio.run(main())
except asyncio.TimeoutError:
    pass

Operational Checklist

Every external call has a timeout.
Define a request/job deadline and enforce it.
Expose timeout metrics and log timeout events with context.

Failure Modes

Hung threads/tasks: missing timeouts cause resource exhaustion.
Timeout mismatch: upstream times out first, leaving downstream still working.

← Retries and Backoff (When Retries Are Harmful)

Idempotency Patterns (Safe Retries) →