error rate
Trips when the ratio of errors to total requests exceeds a threshold.
How it works
Measures failure density. It normalizes error counts against traffic volume, allowing a fixed threshold (e.g., 5%) to remain valid regardless of traffic scale.
When to use
- Detecting degraded service quality in variable-traffic environments.
- Catching partial outages where a service is functional but unstable.
- Monitoring success rate contracts.
When not to use
- Extremely low-traffic endpoints where a single error skews results.
- Binary health checks; use consecutive_failures for those.
State explanations
closed
Failure density is within acceptable bounds for the current traffic volume.
half open
Confirming if the service can handle a limited volume of requests without failure.
open
The proportion of failed requests indicates a systemic issue.