Part 1 of 8 · 1 min read
Why this page comes second
In this module you will
- Explain why every network call needs a timeout, and choose a value from measured response times.
- Write the capped exponential backoff formula and add jitter to it.
- Decide when a failed call should be retried and when it should fail fast.
- Show with numbers how retries at several levels multiply the load.
- Describe where to put the timeout and retry code so no call is missed.
Page 1 showed that a call that never returns drains pools and starts cascades. The timeout is the first defence, and the retry is the first repair. Both are easy to write and easy to get wrong. A bad retry turns a small fault into a big one. The book used is "Understanding Distributed Systems" (UDS), chapter 16. The numbers are made up to show the idea, unless the text names a book.