Axum 0.8 Zero-Copy Streaming and Tower Layer Backpressure Invariants

--- description: Axum 0.8 Zero-Copy Streaming and Tower Layer Backpressure Invariants - Enforces strict backpressure propagation and zero-copy byte slicing across Axum 0.8 routes and Tower middleware layers. Prevents catastrophic memory amplification under high-concurrency I/O by prohibiting unbounded body materialization into memory. Guarantees deterministic O(1) allocation overhead during asynchronous pipeline forwarding. globs: crates/**/src/**/*.rs, src/**/*.rs, **/middleware/**/*.rs, **/handlers/**/*.rs alwaysApply: true verified_by: "agentrules.org" license: "CC-BY-4.0" source: "https://agentrules.org/rules/axum-08-streaming-tower-backpressure-invariants" --- # Axum 0.8 Zero-Copy Streaming and Tower Layer Backpressure Invariants > **Severity**: CRITICAL | **Category**: backend > **Anti-Pattern Warning**: Unbounded Eager Body Materialization and Backpressure Dropping in Tower Middleware ## System Constraints You are a Principal Systems Architect enforcing Axum 0.8 Zero-Copy Streaming and Tower Layer Invariants. Mandate strict backpressure propagation, explicit byte frame budgets, and stream composition without eager heap accumulation. Reject any code that loads arbitrary payloads into memory using unbounded buffers or circumvents the `http_body::Body` frame polling contract. ## Strict Guidelines - Engineers MUST ALWAYS enforce explicit byte limits via `http_body_util::Limited` or `axum::extract::DefaultBodyLimit` before any stream aggregation occurs. - Engineers MUST NOT call `axum::body::to_bytes` with unbounded limits (`usize::MAX`) or collect entire bodies into `Vec<u8>` inside Tower middleware. - Middleware transformations MUST preserve backpressure by wrapping `http_body::Body` implementations directly rather than buffering data through unbounded `mpsc::unbounded_channel` buffers. - Stream payload transformations MUST leverage reference-counted zero-copy slicing (`bytes::Bytes::slice`) instead of deep-cloning byte payloads across layer boundaries. - Handlers handling continuous I/O MUST consume request bodies via `Body::into_data_stream` or `http_body_util::BodyStream` and emit responses using `axum::body::Body::from_stream`. - Custom Tower `Layer` and `Service` implementations MUST propagate `poll_ready` state faithfully to the inner service and not poll downstreams until inner readiness is confirmed. ## Bad vs Good Implementations ### Tower Middleware Body Transformation: Eager In-Memory Buffer vs. Zero-Copy Streaming Wrapper #### Bad Antipattern (AI Common Hallucination) > **Why it fails**: Using `axum::body::to_bytes(body, usize::MAX)` loads the entire request body into heap memory before forwarding it to the inner service. This drops HTTP/2 stream-level backpressure and exposes the server to unbounded memory amplification when concurrent clients upload large payloads. ```rust use axum::{body::Body, extract::Request, response::Response, middleware::Next}; use http_body_util::BodyExt; // ANTI-PATTERN: Eagerly collects entire request body into memory, // breaking backpressure and allowing O(N) memory amplification DOS. pub async fn bad_audit_middleware(request: Request, next: Next) -> Response { let (parts, body) = request.into_parts(); // Materializes unbounded stream into contiguous RAM let bytes = match axum::body::to_bytes(body, usize::MAX).await { Ok(b) => b, Err(_) => return Response::builder().status(400).body(Body::empty()).unwrap(), }; println!("Payload size: {}", bytes.len()); // Reconstructs body from collected bytes, allocating new buffer let new_request = Request::from_parts(parts, Body::from(bytes)); next.run(new_request).await } ``` #### Production Standard (Correct) > **Standard**: Wraps the request body in a zero-allocation `HttpBody` adapter that observes chunk metrics dynamically during stream execution. Downstream backpressure is strictly maintained via `poll_frame` passthrough, chunks are observed without cloning using `frame.data_ref()`, and excessive memory allocation is structurally impossible. ```rust use std::task::{Context, Poll}; use std::pin::Pin; use axum::{body::Body, extract::Request, response::Response, middleware::Next}; use http_body::{Body as HttpBody, Frame}; use bytes::Bytes; use pin_project_lite::pin_project; pin_project! { pub struct InspectStreamBody<B> { #[pin] inner: B, bytes_observed: usize, max_limit: usize, } } impl<B: HttpBody<Data = Bytes>> HttpBody for InspectStreamBody<B> { type Data = Bytes; type Error = B::Error; fn poll_frame( self: Pin<&mut Self>, cx: &mut Context<'_>, ) -> Poll<Option<Result<Frame<Self::Data>, Self::Error>>> { let this = self.project(); match this.inner.poll_frame(cx) { Poll::Ready(Some(Ok(frame))) => { if let Some(data) = frame.data_ref() { *this.bytes_observed += data.len(); // Backpressure invariant: Enforce hard limit during stream traversal if *this.bytes_observed > *this.max_limit { return Poll::Ready(None); } } Poll::Ready(Some(Ok(frame))) } other => other, } } } // PRODUCTION INVARIANT: Zero-copy wrapping preserves downstream backpressure and limits heap usage pub async fn good_audit_middleware(request: Request, next: Next) -> Response { let (parts, body) = request.into_parts(); let wrapped_body = InspectStreamBody { inner: body, bytes_observed: 0, max_limit: 10 * 1024 * 1024, // Strict 10MB budget }; let new_request = Request::from_parts(parts, Body::new(wrapped_body)); next.run(new_request).await } ```

Architectural Specification & Invariant Rationale

Architectural Invariant Design: Axum 0.8 Zero-Copy Streaming and Tower Layer Backpressure Invariants

Axum 0.8 Zero-Copy Streaming and Tower Layer Backpressure Invariants defines an industrial-grade behavioral standard for backend engineering. Enforces strict backpressure propagation and zero-copy byte slicing across Axum 0.8 routes and Tower middleware layers. Prevents catastrophic memory amplification under high-concurrency I/O by prohibiting unbounded body materialization into memory. Guarantees deterministic O(1) allocation overhead during asynchronous pipeline forwarding. Modern software architectures require strict adherence to deterministic patterns to guarantee high availability, maintainability, and zero security vulnerabilities.

Framework Paradigm Shift & Drift MechanicsLLMs are trained on millions of open-source repositories containing outdated code, legacy idioms, and unpatched antipatterns. When coding assistants generate implementations for this stack, they default to common statistical tokens: Unbounded Eager Body Materialization and Backpressure Dropping in Tower Middleware Without an explicit system instruction file, developers spend hours debugging silent failures.

Failure Modes & Production Blast Radius

Hazard Analysis: What Breaks Without This Rule

The following failure modes have been cataloged from hundreds of hours of production code reviews where AI assistants produced non-compliant implementations:

Failure Mode	AI Symptom (The Anti-Pattern)	Severity	Production Impact & Remedy
AI Hallucination & Deprecation Drift	Unbounded Eager Body Materialization and Backpressure Dropping in Tower Middleware	Critical	Violating this industrial standard introduces regression bugs, security risks, and unexpected runtime failures in production. Enforce explicit behavioral system directives across all AI coding assistants in your repository.
Silent Architecture Inconsistency	AI assistants produce mixed paradigms across files, creating fragmented codebase conventions.	Medium	Increases cognitive load during code reviews and leads to maintenance debt across engineering teams. Pin project conventions in .cursor/rules, CLAUDE.md, and AGENTS.md so all team agents follow identical guardrails.

Architectural Directives in Detail

Step-by-Step Production Directives

Each rule constraint below is injected into the AI agent's system prompt to establish deterministic boundaries:

Engineers MUST ALWAYS enforce explicit byte limits via `http_body_util

Engineers MUST ALWAYS enforce explicit byte limits via `http_body_util::Limited` or `axum::extract::DefaultBodyLimit` before any stream aggregation occurs.

Rationale:AI models frequently generate unconstrained code here because statistical language distributions lack project-specific architectural context. Mandating this directive ensures that generated code adheres to verified industrial production requirements.