Exponential Backoff

This is a Go port of the exponential backoff algorithm from Google's HTTP Client Library for Java.

Exponential backoff is an algorithm that uses feedback to multiplicatively decrease the rate of some process, in order to gradually find an acceptable rate. The retries exponentially increase and stop increasing when a certain threshold is met.

Install

go get github.com/cenkalti/backoff/v7

Note the /v7 at the end of the import path.

Usage

For most cases, wrap the operation you want to retry in Retry:

result, err := backoff.Retry(ctx, func() (string, error) {
    resp, err := http.Get("https://www.example.com")
    if err != nil {
        return "", err // transient: Retry will try again
    }
    defer resp.Body.Close()

    switch {
    case resp.StatusCode >= 500:
        return "", fmt.Errorf("server error: %s", resp.Status) // retried
    case resp.StatusCode >= 400:
        // client errors won't fix themselves, so stop retrying.
        return "", backoff.Permanent(fmt.Errorf("client error: %s", resp.Status))
    }
    return "ok", nil
}, backoff.WithMaxTries(5))

Retry runs the operation at least once and keeps retrying with exponential backoff until it succeeds, returns a Permanent error, or a limit is reached. See example_test.go for a fuller example, and the package docs for the available options (WithBackOff, WithMaxTries, WithMaxElapsedTime, WithNotify).

If Retry does not fit your needs, copy it from retry.go and adapt it.

Handling errors

On failure, Retry always returns a *RetryError. It carries the last operation error (LastErr) and the reason retrying stopped (Cause). Inspect it with errors.Is, or reach the struct with AsRetryError:

result, err := backoff.Retry(ctx, operation)
switch {
case errors.Is(err, backoff.ErrPermanent):
    // the operation returned a Permanent error
case errors.Is(err, context.Canceled), errors.Is(err, context.DeadlineExceeded):
    // the caller's context was cancelled or its deadline expired
case errors.Is(err, backoff.ErrMaxElapsedTime):
    // the WithMaxElapsedTime budget was exhausted
case errors.Is(err, backoff.ErrExhausted):
    // WithMaxTries was reached or the backoff policy returned Stop
}

// The last operation error is always available, whatever the cause:
if re := backoff.AsRetryError(err); re != nil {
    log.Printf("gave up after last error: %v", re.LastErr)
}

Mark an error non-retriable with backoff.Permanent(err); Retry stops immediately and returns a *RetryError whose Cause is ErrPermanent and whose LastErr is err.

Bounding total time

Two independent limits cap how long Retry runs, and they behave differently:

• A context deadline (context.WithTimeout) is reactive: it interrupts the wait between attempts and — if your operation observes the context — can abort an in-flight attempt. Retry reports it as context.DeadlineExceeded.
• WithMaxElapsedTime bounds only retry scheduling: it is checked between attempts, never interrupts a running operation, and is reported as ErrMaxElapsedTime.

WithMaxElapsedTime defaults to 15 minutes, so both limits are active unless you override it — pass backoff.WithMaxElapsedTime(0) to rely solely on the context.

Contributing

• I would like to keep this library as small as possible.
• Please don't send a PR without opening an issue and discussing it first.
• If proposed change is not a common use case, I will probably not accept it.