Preventing Failure Propagation with CircuitBreaker

To prevent failures occurring during inter-service communication from spreading to the entire system, CircuitBreaker is introduced. Like a home circuit breaker, it detects the point where a problem occurs and blocks failing requests, helping the system recover and reducing user wait time.

CircuitBreaker has three states: Closed (normal), Open (failure), and Half Open (recovery assessment). When the failure rate or slow call rate reaches a threshold, it switches to the Open state and blocks external requests; after a certain time passes, it performs test requests in the Half Open state to determine whether things are back to normal.

In the current MSA environment, using the lightweight Java library Resilience4j is recommended instead of the existing Netflix Hystrix. In addition to CircuitBreaker, Resilience4j provides various modules such as Retry, RateLimiter, TimeLimiter, Bulkhead, and Cache, and each module is applied in layers according to a set priority order.