single point of failure
termnetworkingops culture
A component whose loss stops the service - regardless of how much redundancy surrounds it.
They are usually found by accident because the obvious ones get removed early and the remaining ones are shared dependencies nobody drew: the identity provider in front of both regions, the certificate authority, the one person who knows the procedure, the DNS zone. Redundancy at one layer often hides a single point at another - two data centres on one fibre route, two clusters in one cloud account, two nodes with one power feed - which is why the useful exercise is tracing a dependency rather than counting devices.
Also known as: spof