Fault Domain
A boundary within which a single failure is contained, defined by the infrastructure and dependencies that components inside it share.
A fault domain is the set of things that fail together. Identifying them correctly is what turns redundancy from an assumption into a property.
Physical and platform domains: host, rack (shared power and top-of-rack switch), availability zone (shared building, power and cooling), region.
Logical domains, which are more often missed: a deployment unit — everything running the same version fails together on a bad release; a configuration scope — everything reading the same configuration key; a shared dependency — every service using one identity provider or one database cluster; a control plane.
The design instruction is simple to state and hard to follow: redundant components must sit in different fault domains, across every relevant dimension. Three replicas in three zones are protected against zone loss and not against a bad deployment, unless the deployment is also staged across zones.
The most valuable output of mapping domains is finding the ones with no redundancy at all — typically a global configuration store, a single DNS zone or one identity provider — because those are the components whose failure is total, and they are rarely where the resilience effort has gone.