std/backoff

Index

• type Backoff
  • method Backoff.delay
• fn retry
• fn retry_notify
• fn always

Types

Backoff

struct Backoff {
  initial: Duration = 500ms
  multiplier: Float = 1.5
  randomization: Float = 0.5
  max_interval: Duration = 60s
  max_elapsed: Duration = 15m
  max_tries: Int = 0
}

The schedule (cenkalti/backoff v4’s defaults): the first delay is initial, each next one multiplier times the last up to max_interval, each randomized within randomization of itself.

Source: std/backoff.xo:10

Backoff.delay

fn (b: Backoff) delay(attempt: Int, rand: Rand) -> Duration uses rand

The delay before retry number attempt (0 for the first retry): initial times multiplier attempt times, capped at max_interval, randomized with one rand.float().

Source: std/backoff.xo:26

Functions

retry

fn retry[T, E, fx: Effects](b: Backoff, clock: Clock, rand: Rand, op: fn() -> T ! E uses fx, retryable: fn(E) -> Bool) -> T ! E uses (clock, rand, fx)

Calls op until it succeeds, retryable(e) is false, or a limit of b runs out, and then answers op’s last result; sleeps on clock between tries (a cancellation point: a cancelled task unwinds there).

Source: std/backoff.xo:31

retry_notify

fn retry_notify[T, E, fx: Effects, rx: Effects, nx: Effects](b: Backoff, clock: Clock, rand: Rand, op: fn() -> T ! E uses fx, retryable: fn(E) -> Bool uses rx, notify: fn(E, Duration) -> () uses nx) -> T ! E uses (clock, rand, fx, rx, nx)

retry, calling notify(e, d) with each error before sleeping d (cenkalti/backoff v4’s RetryNotify): to log or count the retries. retryable and notify may have effects of their own (each its own effect variable: op’s need not include a log’s).

Source: std/backoff.xo:37

always

fn always[E]() -> fn(E) -> Bool

A retryable that retries every error.

Source: std/backoff.xo:40