CircuitBreakerOptionsType
Este conteúdo não está disponível em sua língua ainda.
CircuitBreakerOptionsType =
object
Defined in: src/pattern/CircuitBreakerOptions.ts:65
Plain options-object shape accepted by a CircuitBreaker.
Properties
Section titled “Properties”backoffFactor?
Section titled “backoffFactor?”
readonlyoptionalbackoffFactor?:number
Defined in: src/pattern/CircuitBreakerOptions.ts:103
Multiplier applied to the reopen window per consecutive open, >= 1.
Defaults to DEFAULT_CIRCUIT_BREAKER_BACKOFF_FACTOR (1 — no
growth, the shipped behaviour).
The window is resetTimeoutMs * backoffFactor^(consecutive opens - 1),
clamped to maxResetTimeoutMs. Below 1 it would shrink, so a
breaker would probe a dependency more often the longer that dependency
stayed broken — the inverse of what a backoff is for, which is why the
validator refuses it rather than treating it as an exotic choice.
callTimeoutMs?
Section titled “callTimeoutMs?”
readonlyoptionalcallTimeoutMs?:number
Defined in: src/pattern/CircuitBreakerOptions.ts:71
Per-call timeout; exceeding this counts as a failure.
ignoredErrorNames?
Section titled “ignoredErrorNames?”
readonlyoptionalignoredErrorNames?: readonlystring[]
Defined in: src/pattern/CircuitBreakerOptions.ts:130
Error.name values that never count as a failure. Absent and [] mean
the same thing, which is what lets the HOCON leaf publish [].
Checked before isFailure, and that order is the point: this
list is the operator’s half of the classifier and arrives from a config
file, while isFailure is the developer’s and is compiled in. If the
predicate could override the list, an operator adding a name would be
writing a key that goes inert for exactly the deployments that needed it.
"CircuitBreakerTimeoutError" is listable and means “a call that blew its
own callTimeoutMs does not count against the breaker” — coherent, and
surprising enough to be worth saying out loud.
isFailure?
Section titled “isFailure?”
readonlyoptionalisFailure?: (err) =>boolean
Defined in: src/pattern/CircuitBreakerOptions.ts:80
Optional: classify errors as non-failures to bypass breaker counting.
Belongs to the instance, so it is right for a breaker built in code and
wrong for one resolved by a shared id — there the first caller’s predicate
is the only one that ever runs. CircuitBreaker.call takes the same
predicate per call for that case, and it wins over this one (#874).
Parameters
Section titled “Parameters”Error
Returns
Section titled “Returns”boolean
maxFailures
Section titled “maxFailures”
readonlymaxFailures:number
Defined in: src/pattern/CircuitBreakerOptions.ts:67
Consecutive failures before the breaker opens. Must be >= 1.
maxResetTimeoutMs?
Section titled “maxResetTimeoutMs?”
readonlyoptionalmaxResetTimeoutMs?:number
Defined in: src/pattern/CircuitBreakerOptions.ts:91
Ceiling the reopen window may grow to under backoffFactor, in ms. Defaults to DEFAULT_CIRCUIT_BREAKER_MAX_RESET_TIMEOUT_MS.
It bounds the whole window rather than only the growth, so a
resetTimeoutMs above the ceiling is a contradiction rather than a
silently shortened window — the validator refuses that pair and names this
field. Clamping quietly would take a reopen window an operator chose and
replace it with one nobody wrote down.
random?
Section titled “random?”
readonlyoptionalrandom?: () =>number
Defined in: src/pattern/CircuitBreakerOptions.ts:138
Override Math.random for deterministic tests. No HOCON leaf and it
cannot have one — a config file expresses values, not functions. The same
seam RetryOptions.random and BrokerCommonOptionsType.reconnect.random
carry, and without it the jittered schedule is only assertable through a
range, which is a test that passes for a broken multiplier.
Returns
Section titled “Returns”number
randomFactor?
Section titled “randomFactor?”
readonlyoptionalrandomFactor?:number
Defined in: src/pattern/CircuitBreakerOptions.ts:115
Jitter fraction in [0, 1] applied to the reopen window: the delay is
multiplied by 1 + random(-randomFactor, +randomFactor). Defaults to
DEFAULT_CIRCUIT_BREAKER_RANDOM_FACTOR (0 — no jitter).
Same contract as RetryOptions.randomFactor and
ExponentialBackoffOptions.randomFactor. What it buys a breaker is the
same thing it buys a reconnect loop: N replicas that opened on one
upstream event stop probing it in lockstep, so the recovering dependency
meets a spread of probes instead of a synchronised wave.
resetTimeoutMs
Section titled “resetTimeoutMs”
readonlyresetTimeoutMs:number
Defined in: src/pattern/CircuitBreakerOptions.ts:69
How long the breaker stays open before letting a probe through. ms.
scheduler?
Section titled “scheduler?”
readonlyoptionalscheduler?:Scheduler
Defined in: src/pattern/CircuitBreakerOptions.ts:148
Where the breaker reads the time and arms its call timeout. Default: the wall clock.
The reset window is the reason this exists. A realistic one is seconds, so
a test of “the breaker reopens after the window” either waits those seconds
or shortens the window until it is asserting a shape production never takes.
Pass system.scheduler and a ManualScheduler crosses it for free (#1424).
