add support for retrying 429 (too many requests) responses #45
Conversation
src/config.rs
Outdated
| /// Retry requests with a successful response of 429 (too many requests). | ||
| struct Retry429; | ||
| impl RetryableStrategy for Retry429 { | ||
| fn handle(&self, res: &Result<Response, reqwest_middleware::Error>) -> Option<Retryable> { | ||
| match res { | ||
| // Retry if response status is 429 | ||
| Ok(success) if success.status() == 429 => Some(Retryable::Transient), | ||
| // Otherwise do not retry a successful request | ||
| Ok(_) => None, | ||
| // Retry failures due to network errors | ||
| Err(error) => default_on_request_failure(error), |
There was a problem hiding this comment.
based on the default strategy, I'm wondering if there are same Retryable::Fatal cases not being caught?
https://docs.rs/reqwest-retry/latest/src/reqwest_retry/retryable_strategy.rs.html#112
There was a problem hiding this comment.
I hadn't read the default strategy well enough. I've reverted to using the default strategy instead of specifying a new one. I see the default strategy handles the 429 status code and some other cases.
src/client.rs
Outdated
| @@ -39,7 +40,7 @@ pub mod taxes; | |||
| /// [`Arc`]: std::sync::Arc | |||
| #[derive(Debug)] | |||
| pub struct Client { | |||
| pub(crate) inner: reqwest::Client, | |||
| pub(crate) client_retryable: ClientWithMiddleware, | |||
There was a problem hiding this comment.
I think you could keep this named inner? The reason this was changed in https://github.com/MaterializeInc/rust-frontegg/pull/13/files was because there was no true inner client, we selectively chose one of two clients depending on the http request method.
There was a problem hiding this comment.
OK that makes sense. I've reverted back to using inner.
| } | ||
| } | ||
| } | ||
|
|
||
| impl ClientBuilder { | ||
| /// Sets the policy for retrying failed API calls. | ||
| /// | ||
| /// Note that the created [`Client`] will retry all API calls that return a 429 status code. |
There was a problem hiding this comment.
If we're using the default policy I don't believe this is entirely accurate, it'll also retry server timeouts
Motivation
Adds support for retrying requests when orb responds with a 429 status code. https://docs.withorb.com/reference/rate-limits
Checklist
[Unreleased]section header.