Stopping Request Execution

To intercept and stop a request, return a Response object from your on_request wrapped functions:

Response::new(401)
    .with_headers(vec![("WWW-Authenticate".to_string(), "Bearer realm=\"oauth2\"".to_string())])
    .with_body(r#"{ "error": "token was not present"}"#)

You can block or allow requests to reach the upstream service using a Flow. A Flow in an enum value with two possible values:

Continue: Defines an object that is forwarded to the response.
Break: Aborts the request and returns the provided response.

Create a Flow as follows:

async fn request_filter(request_state: RequestState) -> Flow<()> {
    let header_state = request_state.into_headers_state().await;
    let handler = header_state.handler();
    if handler.header("authorization").is_some() {
        Flow::Continue(())
    } else {
        Flow::Break(Response::new(401)
        .with_headers(vec![("WWW-Authenticate".to_string(), "Bearer realm=\"oauth2\"".to_string())])
        .with_body(r#"{ "error": "token was not present"}"#))
    }
}

Due to the streaming nature of proxy wasm, it is possible for an early request to partially reach the upstream service while awaiting the headers_state.into_body_state function. Avoid this is by not awaiting the body. However, this is not possible for all policies. Configure your upstream to ignore partial requests if the policy must read the body.

To avoid early responses reaching the client, see Sharing Data Between Requests and Responses.

Stopping Request Execution

See Also