deadline propagation
Passing a request's absolute deadline to every service it calls, so each knows how much time is left.
In more detail
Giving each hop a fresh timeout lets downstream work outlive the user. With propagation each service computes its remaining budget, stops when it is spent and records it in traces.
Example
rid=7fa2 with budget 800 ms at the gateway, 640 ms left by the queue, 210 ms left at service B.Related terms
- deadline — The point in time by which a request must finish, after which its answer is no longer useful.
- cancellation — Stopping work that is no longer needed, such as a search request made obsolete by a new keystroke.
- timeout — A limit on how long a caller waits for an answer before giving up.
Taught in
- Timeouts, And Avoiding Cascaded Failure — Scalable System Design & Distributed Architecture