Waffle House has a hurricane disaster plan. It has everything that you want from your IT disaster plans, including contact trees, failover states, and runbooks on partial operation. I think if we took some time to look at how other industries handle outages, we could construct better state machines about our own up/down conditions. I also think that we could use status instrumentation to trigger feature flags that route people to different versions of our products, depending on status.
This talk explores lessons about state that we can adapt from the world outside computers, how to quantify them using a finite state machine, and then how to explore implementing them automatically while we are in a less-than-perfect condition.
Audiences will leave with a new way to think about resilience in the face of downtime and the shades of down that can still be operational. They may also depart with a deep need for smothered hashbrowns.