Timeouts, Retries and Idempotency In Distributed Systems

The definition of insanity is doing the same thing over and over again” - this quote attributed to Einstein warns us of the danger of magical thinking, hoping that trying something just one more time will achieve success when before we failed. But is this really insanity?

In this talk, I’ll argue that retrying things actually does make a lot of sense, and is in fact key to improving the resilience of a distributed system. Along the way, I’ll explain the importance of timeouts, retry limits and knowing when giving up does make sense. I’ll also show how retries can be made safe (and help avoid draining your bank account), and perhaps we’ll get to examine that Einstein quote in a bit more detail.