Yann Simon

I'm trying this PR, and it changes the behavior when using `.layer(DefaultBodyLimit::max(...))`. The requests are not rejected anymore. Edit: Instead of using `req.into_body()`, we should use `req.into_limited_body()`.

When trying out this change with production-like load, I can confirm that the CPU usage issue is fixed.

This change seems to be sufficient to handle limits on body size: https://github.com/tokio-rs/axum/pull/2592

To fix the error in CI: https://github.com/tokio-rs/axum/pull/2596

@jplatte is there anything I can do to help you on this PR?

> cc @yanns @mladedav @ibraheemdev do any of you have ideas / opinions on this? I have never used `nest` myself, having the feeling (maybe wrong) that it can be...

Can you check again https://github.com/Ptrskay3/axum-prometheus? I'm using since years, and it's using `metrics`, and does not force using `localhost:9000`. You put the `route("/metrics", get(|| async move { metric_handle.render() }))` where...

@jplatte I've fixed the merge conflicts from `main` and fixed the compilation errors. Do you think it makes sense to spend time on this? Is it the right direction?

I'm continuing the work in https://github.com/tokio-rs/axum/pull/2725 Feedback, review, help welcome

@jmg-duarte if you're motivated to fix this issue, you're welcome trying. Even a PR with a test reproducing the issue can help!