Skip to content

req_retry() alters req_perform() so that it will automatically retry in the case of failure. To activate it, you must specify either the total number of requests to make with max_tries or the total amount of time to spend with max_seconds. Then req_perform() will retry if:

  • The either the HTTP request or HTTP response doesn't complete successfully leading to an error from curl, the lower-level library that httr uses to perform HTTP request. This occurs, for example, if your wifi is down.

  • The error is "transient", i.e. it's an HTTP error that can be resolved by waiting. By default, 429 and 503 statuses are treated as transient, but if the API you are wrapping has other transient status codes (or conveys transient-ness with some other property of the response), you can override the default with is_transient.

It's a bad idea to immediately retry a request, so req_perform() will wait a little before trying again:

  • If the response contains the Retry-After header, httr2 will wait the amount of time it specifies. If the API you are wrapping conveys this information with a different header (or other property of the response) you can override the default behaviour with retry_after.

  • Otherwise, httr2 will use "truncated exponential backoff with full jitter", i.e. it will wait a random amount of time between one second and 2 ^ tries seconds, capped to at most 60 seconds. In other words, it waits runif(1, 1, 2) seconds after the first failure, runif(1, 1, 4) after the second, runif(1, 1, 8) after the third, and so on. If you'd prefer a different strategy, you can override the default with backoff.


  max_tries = NULL,
  max_seconds = NULL,
  is_transient = NULL,
  backoff = NULL,
  after = NULL



A request.

max_tries, max_seconds

Cap the maximum number of attempts with max_tries or the total elapsed time from the first request with max_seconds. If neither option is supplied (the default), req_perform() will not retry.


A predicate function that takes a single argument (the response) and returns TRUE or FALSE specifying whether or not the response represents a transient error.


A function that takes a single argument (the number of failed attempts so far) and returns the number of seconds to wait.


A function that takes a single argument (the response) and returns either a number of seconds to wait or NULL, which indicates that a precise wait time is not available that the backoff strategy should be used instead..


A modified HTTP request.

See also

req_throttle() if the API has a rate-limit but doesn't expose the limits in the response.


# google APIs assume that a 500 is also a transient error
request("") %>%
  req_retry(is_transient = ~ resp_status(.x) %in% c(429, 500, 503))
#> <httr2_request>
#> GET
#> Body: empty
#> Policies:
#>retry_is_transient: a <rlang_lambda_function/function> object

# use a constant 10s delay after every failure
request("") %>%
  req_retry(backoff = ~ 10)
#> <httr2_request>
#> GET
#> Body: empty
#> Policies:
#>retry_backoff: a <rlang_lambda_function/function> object

# When rate-limited, GitHub's API returns a 403 with
# `X-RateLimit-Remaining: 0` and an Unix time stored in the
# `X-RateLimit-Reset` header. This takes a bit more work to handle:
github_is_transient <- function(resp) {
  resp_status(resp) == 403 &&
    identical(resp_header(resp, "X-RateLimit-Remaining"), "0")
github_after <- function(resp) {
  time <- as.numeric(resp_header(resp, "X-RateLimit-Reset"))
  time - unclass(Sys.time())
request("") %>%
    is_transient = github_is_transient,
    after = github_after
#> <httr2_request>
#> GET
#> Body: empty
#> Policies:
#>retry_is_transient: a function
#>retry_after: a function