Skip to content

Retries

Every Paws service client retries failed requests automatically — this isn't opt-in, and there's no separate retry package or handler to attach; it's built into the shared request pipeline in paws.common that every paws.<category> service uses (see architecture.md).

Default behavior

By default, a request is retried up to 3 times (4 attempts total) before Paws gives up and raises the error to your R code.

library(paws)

svc <- s3(region = "us-west-2")

# If this fails with a retryable error, Paws retries up to 3 times
# with backoff before raising an error to you.
svc$get_object(Bucket = "my-bucket", Key = "file.txt")

Configuring the number of retries

Set max_retries in the service config to change how many retries are attempted. Setting it to 0 disables retries entirely — the first error is raised immediately.

svc <- s3(config(max_retries = 10))

# Fail fast: raise the first error without retrying.
svc <- s3(config(max_retries = 0))

What gets retried

A failed request is retried if either:

  • The error code returned by AWS is a known transient or throttling error, e.g. Throttling, ThrottlingException, ProvisionedThroughputExceededException, RequestTimeout, SlowDown, RequestLimitExceeded, or similar.
  • The HTTP response status code is 500, 502, 503, or 504.

Any other error (e.g. AccessDenied, NoSuchKey, a 4xx validation error) is raised immediately without being retried.

Backoff

Between retries, Paws waits with an exponential backoff: on retry attempt i, it sleeps a random amount of time between 0 and 2^i seconds, capped at a maximum of 20 seconds, before trying again.