Fix dead CircuitBreaker.ResetTimeout config (round 39)

error_recovery.go: ResetTimeout was copied into the CircuitBreaker but never
consulted by the state machine (used only for a metric), so the documented
"how long to wait before fully closing the circuit" (reset_timeout, default
30s) had no effect: Open->HalfOpen used Timeout (60s), and a single
half-open success closed the circuit immediately, losing the intended
open -> half-open -> (reset cooldown) -> closed hysteresis and letting a
stray success immediately reopen (bounce).

Now the circuit remains half-open after entering it until resetTimeout has
elapsed; only then can a successful request fully close it. Half-open
requests are still allowed as probes. Default behavior is more
conservative (up to Timeout+ResetTimeout before full close).

Tests: updated TestCircuitBreakerHalfOpenTransition to the corrected
semantics; added TestCircuitBreakerResetTimeoutEnforced (fails on the old
dead-field code, where resetTimeout was ignored).

Verified: full source-order suite, full -race -count=1 -shuffle=on ./...
(all packages), golangci-lint run ./... (0 issues).
This commit is contained in:
2026-08-13 05:50:33 +01:00
parent 0358d9b970
commit 0d5354bedc
2 changed files with 56 additions and 10 deletions
+15 -1
View File
@@ -191,12 +191,16 @@ type CircuitBreaker struct {
maxFailures int
// timeout is how long to wait before allowing requests in half-open state
timeout time.Duration
// resetTimeout is how long to wait before transitioning from open to half-open
// resetTimeout is how long to stay in half-open (after the open timeout)
// before a successful response fully closes the circuit
resetTimeout time.Duration
// state tracks the current circuit breaker state
state CircuitBreakerState
// failures counts consecutive failures
failures int64
// halfOpenSince is when the circuit entered half-open; used to enforce
// resetTimeout before a success can fully close it
halfOpenSince time.Time
}
// CircuitBreakerConfig holds configuration parameters for circuit breakers.
@@ -275,6 +279,7 @@ func (cb *CircuitBreaker) allowRequest() bool {
case CircuitBreakerOpen:
if now.Sub(cb.lastFailureTime) > cb.timeout {
cb.state = CircuitBreakerHalfOpen
cb.halfOpenSince = now
cb.logger.Infof("Circuit breaker transitioning to half-open state")
return true
}
@@ -317,6 +322,15 @@ func (cb *CircuitBreaker) recordSuccess() {
switch cb.state {
case CircuitBreakerHalfOpen:
// Honor resetTimeout: do not fully close until the circuit has been
// in half-open for at least resetTimeout. This provides the
// documented open -> half-open -> (reset cooldown) -> closed
// hysteresis instead of closing on the very first half-open
// success, which would let a stray success immediately reopen
// (bounce). While in half-open requests are still allowed as probes.
if time.Since(cb.halfOpenSince) < cb.resetTimeout {
return
}
cb.failures = 0
cb.state = CircuitBreakerClosed
cb.LogInfo("Circuit breaker closed after successful request in half-open state")