Stop spending resources on a known unhealthy dependency#
Closed requests flow normally. After a configured failure policy is exceeded, Open calls fail fast or use a fallback. After a cooldown, Half-Open admits limited probes; success closes the breaker and failure opens it again.
| Retry | Circuit breaker |
|---|---|
| Could this request succeed if attempted again? | Should callers currently attempt this dependency? |
| Consumes another attempt | Fails fast while open |
They can work together. Poor thresholds can open on noise; simultaneous probes can spike recovery; stale fallbacks can hide prolonged failure. A breaker contains propagation - it does not repair the dependency.
Related practice problems
Use the manifest-backed prompts below to apply this lesson in a complete design.
- Circuit Breaker patternMicrosoft Azure Architecture Center
- Timeouts, retries, and backoff with jitterAmazon Builders' Library