Expand description
The replica dispatcher — the harness’s stand-in for the k8s Service.
Holds one [Store] handle per gonzalod replica, round-robins each op
(get, get_raw, put, delete) to a replica, and fails over on a
Backend error (transport, dead replica). Conflict (including a
[PutResult::Conflict], which the caller answers by re-reading and
retrying the RMW), NotFound and Serde are answers from a live replica,
returned unchanged and never failover triggers. This is exactly the
failover path a k8s agent pod relies on when a gonzalod pod dies behind
the Service.
Structs§
- Dispatcher
- Fans record ops across N replica stores with round-robin + failover.