FastAPI Lifespan State Graceful Draining & Signal Invariance

--- description: FastAPI Lifespan State Graceful Draining & Signal Invariance - Enforces deterministic teardown of connection pools and active asynchronous worker tasks within ASGI lifespan contexts during container SIGTERM signals. Prevents abrupt socket termination, unhandled CancelledError leaks, and state corruption across Kubernetes zero-downtime rolling deploys. globs: **/main.py, **/app.py, **/lifespan.py, **/core/**/*.py, **/api/**/*.py alwaysApply: true verified_by: "agentrules.org" license: "CC-BY-4.0" source: "https://agentrules.org/rules/fastapi-lifespan-state-graceful-draining-standard" --- # FastAPI Lifespan State Graceful Draining & Signal Invariance > **Severity**: CRITICAL | **Category**: backend > **Anti-Pattern Warning**: Premature Resource Destruction and Fire-and-Forget Task Severance in Deprecated Event Handlers ## System Constraints You are a Principal Backend Reliability Architect enforcing strict ASGI Lifespan protocol conformance and graceful termination invariant guarantees in FastAPI services. You must reject deprecated event handlers, untracked background coroutines, and immediate resource disposal without drain timeouts. ## Strict Guidelines - Engineers MUST ALWAYS use the ASGI lifespan async context manager pattern via `@asynccontextmanager async def lifespan(app: FastAPI)` instead of legacy `@app.on_event('startup')` and `@app.on_event('shutdown')`. - Engineers MUST manage background coroutines via a shared, thread-safe `set[asyncio.Task]` registry stored in `app.state` to enable graceful awaiting during shutdown rather than invoking fire-and-forget `asyncio.create_task` without retention. - Engineers MUST NOT tear down downstream connection pools (e.g., PostgreSQL asyncpg pools, Redis clusters, gRPC channels) before setting the readiness sentinel to False and draining in-flight async tasks. - Code MUST wrap lifespan drain sequences in `asyncio.wait_for()` with a bounded timeout strictly lower than the container orchestrator's `terminationGracePeriodSeconds` (defaulting to a 10-25 second window). - Services MUST catch `asyncio.CancelledError` within tracked background loops and execute clean rollback or state commit sequences prior to yielding final cancellation. - Components NEVER instantiate unmanaged global state or initialize persistent network clients outside the explicit lifespan context generator. ## Bad vs Good Implementations ### FastAPI ASGI Lifespan Connection Pool and Task Draining #### Bad Antipattern (AI Common Hallucination) > **Why it fails**: Using deprecated `@app.on_event` handlers couples resource lifetimes directly to process signals without coordination. Downstream database connections close immediately while in-flight HTTP requests are still executing, and untracked background tasks launched with `asyncio.create_task` are dropped mid-execution, causing silent data loss. ```python from fastapi import FastAPI import asyncio import asyncpg app = FastAPI() db_pool = None @app.on_event("startup") async def startup(): global db_pool db_pool = await asyncpg.create_pool("postgresql://user:pass@db:5432/production") @app.on_event("shutdown") async def shutdown(): # ANTI-PATTERN: Immediatley terminates pool while in-flight requests may still be querying # Background tasks running via create_task are abruptly aborted by SIGKILL without draining await db_pool.close() @app.post("/process") async def process(): # Untracked fire-and-forget task severed mid-flight upon container shutdown asyncio.create_task(db_pool.execute("INSERT INTO audit_log VALUES (NOW())")) return {"status": "accepted"} ``` #### Production Standard (Correct) > **Standard**: Adheres to the official ASGI lifespan protocol. The state object isolates the connection pool and a monitored task registry. Upon SIGTERM, readiness is immediately set to False so orchestration traffic shifts, followed by a bounded wait (`asyncio.wait_for`) allowing in-flight background operations to finish before database pools are closed. ```python from contextlib import asynccontextmanager import asyncio import logging from typing import AsyncGenerator from fastapi import FastAPI, Request import asyncpg logger = logging.getLogger("app.lifespan") @asynccontextmanager async def lifespan(app: FastAPI) -> AsyncGenerator[None, None]: # Phase 1: Resource Initialization pool = await asyncpg.create_pool("postgresql://user:pass@db:5432/production", min_size=5, max_size=20) app.state.db_pool = pool app.state.background_tasks: set[asyncio.Task] = set() app.state.is_ready = True yield # Phase 2: Graceful Draining Sequence app.state.is_ready = False logger.info("Readiness flipped to false. Draining active background tasks...") active_tasks = [t for t in app.state.background_tasks if not t.done()] if active_tasks: try: # Bounded drain: Wait up to 15s for active tasks to conclude await asyncio.wait_for(asyncio.gather(*active_tasks, return_exceptions=True), timeout=15.0) except asyncio.TimeoutError: logger.warning("Timed out waiting for tasks to drain. Cancelling remaining tasks.") for task in active_tasks: task.cancel() await asyncio.gather(*active_tasks, return_exceptions=True) # Phase 3: Resource Teardown logger.info("Closing database connection pool...") await pool.close() app = FastAPI(lifespan=lifespan) @app.post("/process") async def process(request: Request): async def tracked_worker(pool: asyncpg.Pool): try: await pool.execute("INSERT INTO audit_log VALUES (NOW())") except asyncio.CancelledError: logger.warning("Task received cancellation during shutdown drain.") raise task = asyncio.create_task(tracked_worker(request.app.state.db_pool)) request.app.state.background_tasks.add(task) task.add_done_callback(request.app.state.background_tasks.discard) return {"status": "accepted"} ```

Architectural Specification & Invariant Rationale

Architectural Invariant Design: FastAPI Lifespan State Graceful Draining & Signal Invariance

FastAPI Lifespan State Graceful Draining & Signal Invariance defines an industrial-grade behavioral standard for backend engineering. Enforces deterministic teardown of connection pools and active asynchronous worker tasks within ASGI lifespan contexts during container SIGTERM signals. Prevents abrupt socket termination, unhandled CancelledError leaks, and state corruption across Kubernetes zero-downtime rolling deploys. 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: Premature Resource Destruction and Fire-and-Forget Task Severance in Deprecated Event Handlers 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	Premature Resource Destruction and Fire-and-Forget Task Severance in Deprecated Event Handlers	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 use the ASGI lifespan async context manager pattern via `@asynccontextmanager async def lifespan(app

Engineers MUST ALWAYS use the ASGI lifespan async context manager pattern via `@asynccontextmanager async def lifespan(app: FastAPI)` instead of legacy `@app.on_event('startup')` and `@app.on_event('shutdown')`.

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.