Axum WebTransport Datagram Ingress Flow-Control & Backpressure Boundary

--- description: Axum WebTransport Datagram Ingress Flow-Control & Backpressure Boundary - Enforces deterministic ingress backpressure and bounded allocation on WebTransport UDP datagram streams in Axum/Tokio services. Because HTTP/3 datagrams bypass connection-level flow control, unconstrained ingress loops saturate executor thread pools and trigger unbounded memory allocations during packet floods. globs: src/**/*.rs, crates/**/*.rs alwaysApply: true verified_by: "agentrules.org" license: "CC-BY-4.0" source: "https://agentrules.org/rules/axum-webtransport-datagram-flow-control-boundary" --- # Axum WebTransport Datagram Ingress Flow-Control & Backpressure Boundary > **Severity**: CRITICAL | **Category**: backend > **Anti-Pattern Warning**: Unbounded Task Spawning and Unchecked Ingress Loops on HTTP/3 Datagram Channels ## System Constraints You are a Principal Systems Engineer enforcing strict UDP and WebTransport datagram flow-control invariants in Axum and Tokio environments. Mandate zero unbounded allocations, bounded concurrency buffers, and immediate drop-or-throttle backpressure boundaries to prevent thread pool exhaustion and out-of-memory denial-of-service conditions. ## Strict Guidelines - Engineers MUST ALWAYS enforce bounded ingress buffering using bounded channels (e.g., `tokio::sync::mpsc::channel`) or bounded semaphore permits when reading WebTransport datagrams. - Engineers MUST NEVER invoke `tokio::spawn` per incoming datagram directly from the ingress read loop. - Engineers MUST handle Datagram Buffer Saturation deterministically using `try_reserve` or non-blocking drops, emitting metric counters for dropped datagrams instead of stalling the network loop. - Code SHOULD pre-allocate datagram memory slabs using reusable byte pools (e.g., `bytes::BytesMut`) rather than performing continuous heap allocations inside packet-processing hot paths. - Services MUST bound the total memory allocated to datagram ingress per session and across the entire Axum HTTP/3 server process using atomic limits or token buckets. - Components NEVER bridge raw datagrams directly into blocking sync code or unthrottled downstream message buses without strict rate-limiting and circuit breaking. ## Bad vs Good Implementations ### Unbounded Datagram Spawn vs Bounded Permit Reserve #### Bad Antipattern (AI Common Hallucination) > **Why it fails**: Directly spawning asynchronous tasks within an unthrottled datagram loop causes unbounded allocation. WebTransport datagrams have no QUIC flow-control window, making the process vulnerable to executor starvation and memory exhaustion under packet-flood conditions. ```rust use axum::extract::State; use bytes::Bytes; // Anti-pattern: Unbounded task creation and unbounded channel consumption per datagram pub async fn handle_session(mut session: wtransport::Session) { while let Ok(datagram) = session.receive_datagram().await { // Spawns a Tokio task for every UDP packet received. // A flood of 500k pps exhausts the Tokio worker thread queue, // causing catastrophic memory bloat and eventual process OOM. tokio::spawn(async move { let payload = datagram.to_vec(); // Unnecessary allocation process_packet(&payload).await; }); } } async fn process_packet(_data: &[u8]) { // processing logic } ``` #### Production Standard (Correct) > **Standard**: Implements a strict bounded channel with zero-copy packet ingestion and deterministic non-blocking drops (`try_send`) upon buffer saturation. This preserves Tokio worker health and satisfies the flow-control boundary for unreliable datagrams. ```rust use bytes::Bytes; use tokio::sync::mpsc::{channel, error::TrySendError}; use tracing::warn; const DATAGRAM_CAPACITY: usize = 4096; pub async fn handle_session(mut session: wtransport::Session) { // Invariant: Hard-bounded channel enforcing flow-control boundary let (tx, mut rx) = channel::<Bytes>(DATAGRAM_CAPACITY); // Worker pool reader with dedicated bounded capacity tokio::spawn(async move { while let Some(datagram) = rx.recv().await { process_packet(datagram).await; } }); // Ingress loop: Non-blocking try_reserve drop semantics while let Ok(datagram) = session.receive_datagram().await { let payload = datagram.payload(); // Zero-copy Bytes view match tx.try_send(payload) { Ok(_) => {} Err(TrySendError::Full(_)) => { // Drop packet intentionally to preserve executor responsiveness metrics::counter!("webtransport.datagrams.dropped_capacity").increment(1); } Err(TrySendError::Closed(_)) => { warn!("Datagram processor terminated; breaking ingress loop"); break; } } } } async fn process_packet(_data: Bytes) { // processing logic } ```

Architectural Specification & Invariant Rationale

Architectural Invariant Design: Axum WebTransport Datagram Ingress Flow-Control & Backpressure Boundary

Axum WebTransport Datagram Ingress Flow-Control & Backpressure Boundary defines an industrial-grade behavioral standard for backend engineering. Enforces deterministic ingress backpressure and bounded allocation on WebTransport UDP datagram streams in Axum/Tokio services. Because HTTP/3 datagrams bypass connection-level flow control, unconstrained ingress loops saturate executor thread pools and trigger unbounded memory allocations during packet floods. 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 Task Spawning and Unchecked Ingress Loops on HTTP/3 Datagram Channels 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 Task Spawning and Unchecked Ingress Loops on HTTP/3 Datagram Channels	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 bounded ingress buffering using bounded channels (e.g., `tokio

Engineers MUST ALWAYS enforce bounded ingress buffering using bounded channels (e.g., `tokio::sync::mpsc::channel`) or bounded semaphore permits when reading WebTransport datagrams.

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.