Circuit Breaker Pattern

The circuit breaker pattern is a technique that prevents a failing service from being repeatedly called, allowing it time to recover. This protects the entire system from cascading failures that occur when one slow service causes a chain reaction, exhausting thread pools and timeouts which then bring down everything upstream.

There are three states in a circuit breaker: CLOSED (normal operation), OPEN (service is down), and HALF-OPEN (testing recovery). When the failure threshold is exceeded, the circuit trips to the OPEN state, immediately failing requests without making a network call and returning cached or default responses. After a reset timeout, the circuit enters HALF-OPEN state to test if the service has recovered. If successful, it returns to CLOSED; if not, it remains in OPEN.

Circuit breakers are implemented in various programming languages, such as Java (Resilience4j, Hystrix), Node.js (opossum), Go (sony/gobreaker), Python (pybreaker), and .NET (Polly). They can also be integrated into service meshes like Istio, requiring no code changes.

Using circuit breakers offers several benefits: preventing cascading failures, providing quick responses instead of long timeouts, and enabling graceful degradation with fallback strategies like cached responses or static pages. However, implementing circuit breakers adds complexity, and tuning the failure threshold can be tricky to avoid flapping or slow tripping. Probes during the HALF-OPEN state may still allow some failures to pass through, and stale cached data as fallback can mislead users.

Circuit breakers are particularly useful when calling external APIs, microservices, or synchronous calls with timeout risks. They should be avoided for local in-process functions, async message queues, or when already using a service mesh that handles circuit breaking.