Feature Request - Include timestamp of 1st failure event in CircuitBreakerOnFailureRateExceededEvent

[laksnv]

I'd like the ability to know how long it takes for Circuit Breaker to trip: T2 - T1

T2: Timestamp when failure ratio exceeds threshold
T1: Timestamp of first observed failure

If this data was made available through CircuitBreakerOnFailureRateExceededEvent, it would be perfect for clients to record how long detection took. Let me know what you think.

I couldn't find a similar request from the past; please let me know if I missed any.

[RobWin]

Hi,
this kind of information is currently not available. A CircuitBreaker does not store any timestamps.
From the design perspective we wanted to enable it with Event Sourcing. Clients can consume Events and track any other kind of metric/information which is important for them.

[laksnv]

Thanks for getting back!

Even if clients rely on maintaining a record of failure events, they still will need to maintain a sliding window to determine the first failure in the window. Am I understanding it right?

[RobWin]

That's a good point.
Only option is to consume the CircuitBreakerStateTransitionEvent and then calculate backwards based on the failure rate threshold. But it's cumbersome.
But storing the timestamp in the sliding window would introduce a lot of changes.

[laksnv]

We just need to keep track of timestamp of first reported failure, if any, for every second in the sliding window and include this detail when creating a snapshot. CircuitBreaker.Metrics will expose a method to get details about failure timestamps (whether getting the earliest reported failure timestamp or similar). When publishing rate exceeded events, timestamp detail could be included:

resilience4j/resilience4j-circuitbreaker/src/main/java/io/github/resilience4j/circuitbreaker/internal/CircuitBreakerStateMachine.java

Line 951 in dc9505a

publishCircuitFailureRateExceededEvent(getName(), circuitBreakerMetrics.getFailureRate());

.

What else is needed?

[laksnv]

It may be better to only expose earliest failure timestamp on demand i.e. only when generating error events.