Software Engineering · Distributed Systems
A pattern that detects when a downstream service is failing and stops sending it requests temporarily, instead of letting every caller time out and pile up.
Example: After five consecutive failures to a payment API, the circuit 'opens' and immediately returns an error for the next 30 seconds rather than letting callers wait.
In practice: Named after the electrical component — it trips to protect the broader system. Without it, one slow dependency can cascade failures to everything that depends on it through thread exhaustion.