CircuitBreaker [pattern] v1.0.0

A service that calls a remote dependency wraps the call in a circuit breaker. After a threshold of failures, the breaker trips OPEN and short-circuits subsequent calls — they fail immediately without hitting the network. After a cool-down it goes HALF-OPEN to probe; on success it CLOSES, on failure it re-OPENS. This prevents a cascade where one slow downstream consumes all callers' threads.

Label

Circuit Breaker — fail fast on a downstream that is already failing

Problem

Service A calls Service B synchronously. B becomes slow (5s timeouts). A's request threads pile up waiting on B. Soon every A thread is blocked on B, and A becomes unavailable to all callers. The failure has cascaded.

Solution

    Wrap the remote call in a circuit-breaker library (Resilience4j, Polly, Hystrix-equivalent).

      State machine:
        CLOSED    → calls flow through; failures counted in a sliding window.
        OPEN      → after N failures within W seconds, trip. New calls fail immediately
                    with CircuitBreakerOpenException.
        HALF-OPEN → after cool-down T, allow one probe. Success → CLOSED. Failure → OPEN.

      Pair with a fallback: cached response, default value, or graceful degradation.
  

CircuitBreaker [pattern] v1.0.0

A service that calls a remote dependency wraps the call in a circuit breaker. After a threshold of failures, the breaker trips OPEN and short-circuits subsequent calls — they fail immediately without hitting the network. After a cool-down it goes HALF-OPEN to probe; on success it CLOSES, on failure it re-OPENS. This prevents a cascade where one slow downstream consumes all callers' threads.

Label

Circuit Breaker — fail fast on a downstream that is already failing

Problem

Service A calls Service B synchronously. B becomes slow (5s timeouts). A's request threads pile up waiting on B. Soon every A thread is blocked on B, and A becomes unavailable to all callers. The failure has cascaded.

Solution

    Wrap the remote call in a circuit-breaker library (Resilience4j, Polly, Hystrix-equivalent).

      State machine:
        CLOSED    → calls flow through; failures counted in a sliding window.
        OPEN      → after N failures within W seconds, trip. New calls fail immediately
                    with CircuitBreakerOpenException.
        HALF-OPEN → after cool-down T, allow one probe. Success → CLOSED. Failure → OPEN.

      Pair with a fallback: cached response, default value, or graceful degradation.
  

Relations

enhances: [@backend-architecture/principle-disposability]

Label

Circuit Breaker — fail fast on a downstream that is already failing

Problem

Service A calls Service B synchronously. B becomes slow (5s timeouts). A's request threads pile up waiting on B. Soon every A thread is blocked on B, and A becomes unavailable to all callers. The failure has cascaded.

Solution

    Wrap the remote call in a circuit-breaker library (Resilience4j, Polly, Hystrix-equivalent).

      State machine:
        CLOSED    → calls flow through; failures counted in a sliding window.
        OPEN      → after N failures within W seconds, trip. New calls fail immediately
                    with CircuitBreakerOpenException.
        HALF-OPEN → after cool-down T, allow one probe. Success → CLOSED. Failure → OPEN.

      Pair with a fallback: cached response, default value, or graceful degradation.
  

Enhances

• @backend-architecture/principle-disposability