3-Hour Debugging: How `time.sleep` in Async Functions Kil...

DEV Community

Authentication Security Deep Dive: From Brute Force to Salted Hashing (With Java Examples) Why AI Systems Don’t Fail — They Drift Spilling beans for how i learn for exam😁"Reinforcement Learning Cheat Sheet" I Replaced Chrome with Safari for AI Browser Automation. Here's What Broke (and What Finally Worked) How Python Borrows Other People's Work The $40 Architecture: Processing 1 Billion API Requests with 99.99% Uptime Vibe Coding: A Workflow Guide (From Zero to SaaS) Most webhook security guides protect the wrong side. The scary part is delivery. Headless CMS for TanStack Start: Build a Blog with Cosmic EU Age Verification App "Hacked in 2 Minutes" — What Actually Happened Comfy Cloud’s delete function does not actually remove files Running AI Models on GPU Cloud Servers: A Beginner Guide Event-driven media intelligence with AWS Step Functions and Bedrock I scored 500 AI prompts across 8 quality dimensions — here's what broke How to Call Google Gemini API from Next.js (Free Tier, No Backend Needed) The Portal Protocol: Reclaiming Human Connection in the Age of AI How to Fix Your Team's Scattered Knowledge Problem With a Self-Hosted Forum Intro to tc Cloud Functors: A Graph-First Mental Model for the Modern Cloud Designing Multi-Tenant Backends With Both Ownership and Team Access I Built a Neumorphic CSS Library with 77+ Components — Here's What I Learned PostgreSQL Performance Optimization: Why Connection Pooling Is Critical at Scale Cómo construí un SaaS multi-rubro para gestionar expensas en Argentina con FastAPI + Vue 3 🚀 I Built an Ethical Hacking Scanner Tool – Open Source Project I Replaced /usage and /context in Claude Code With a Single Statusline A Pythonic Way to Handle Emails (IMAP/SMTP) with Auto-Discovery and AI-Ready Design I Collected 8.9 Million Polymarket Price Points — Here's What I Found About How Markets Really Move EcoTrack AI — Carbon Footprint Tracker & Dashboard Everyone's Using AI. No One Agrees How. 5 self-hosted ebook managers worth trying in 2026 Building Your First AI Agent with LangChain: From Chatbot to Autonomous Assistant

BAOFUFAN · 2026-05-01 · via DEV Community

Here’s the situation: we have a data collection service that fetches data from a dozen upstream APIs. The synchronous version took a painful 30 minutes per run. I thought, “This is exactly the kind of problem asyncio was built for!” I spent an afternoon replacing <code>requests</code> with <code>aiohttp</code>, decorating every function with <code>async</code>/<code>await</code>, and ran the code — same 30 minutes. Not a single second faster. I was floored. Eventually, I tracked the culprit to a stray <code>time.sleep(0.5)</code> buried deep inside a nested function. That half-second sleep was enough to freeze the entire event loop inside that coroutine, turning our glorious “async concurrency” back into plain old serial execution. The takeaway? Some of asyncio’s most counterintuitive landmines are impossible to fully appreciate until you step on one yourself. Here’s the full post-mortem: the debugging journey, the root cause, and how to avoid this trap for good. <h2> Why a Single <code>sleep</code> Can Destroy Concurrency </h2> Let’s recap how asyncio works: the event loop is essentially a single-threaded scheduler that manages a queue of tasks. Coroutines defined with <code>async def</code> yield control back to the event loop whenever they <code>await</code>, allowing the loop to switch to other ready coroutines and keep making progress. But yielding must be explicit. <code>await asyncio.sleep(n)</code> registers a timer with the event loop and immediately hands back control — other tasks get their turn. In contrast, <code>time.sleep(n)</code> is a synchronous blocking call. It puts the entire thread to sleep, the event loop gets zero control, and every coroutine you wrote simply waits in line, no matter how many tasks you’ve created. In plain terms: <ul> <li> <code>await asyncio.sleep()</code>: “I’ll set a timer with the event loop and kindly step aside so others can work.”</li> <li> <code>time.sleep()</code>: “I’m going to take a nap, and no one — not even the event loop — can do anything until I wake up.”</li> </ul> <h2> Bad Practice vs. The Right Way </h2> Bad code (looks async but blocks the single thread): <div class="highlight js-code-highlight"> <pre class="highlight python"><code>import asyncio import time async def fetch_data(url: str): # 模拟请求前处理 print(f"开始请求 {url}") time.sleep(0.5) # ❌ 同步阻塞，整个事件循环停滞 # 这里还会去发起 aiohttp 请求等等 print(f"完成请求 {url}") async def main(): urls = [f"https://api.example.com/data/{i}" for i in range(10)] # 看似并发启动 tasks = [asyncio.create_task(fetch_data(url)) for url in urls] await asyncio.gather(*tasks) asyncio.run(main()) </code></pre> </div> When you run this, you’ll see the prints appear one by one in order. Ten tasks take over 5 seconds, and concurrency is a complete illusion. Good code (using <code>asyncio.sleep</code> to yield control): <div class="highlight js-code-highlight"> <pre class="highlight python"><code>import asyncio import aiohttp async def fetch_data(session: aiohttp.ClientSession, url: str) -> dict: """ 真正的异步请求函数：IO 全部交给事件循环调度 """ print(f"开始请求 {url}") # 模拟速率限制等待，使用 asyncio.sleep，不阻塞其他协程 await asyncio.sleep(0.5) async with session.get(url, timeout=10) as resp: data = await resp.json() print(f"完成请求 {url}, 状态码 {resp.status}") return data async def main(): urls = [f"https://api.example.com/data/{i}" for i in range(10)] # 使用连接池复用 TCP 连接，减少开销 connector = aiohttp.TCPConnector(limit=20) # 最大并发连接数 async with aiohttp.ClientSession(connector=connector) as session: tasks = [fetch_data(session, url) for url in urls] results = await asyncio.gather(*tasks, return_exceptions=True) # 简单错误处理 for i, result in enumerate(results): if isinstance(result, Exception): print(f"请求 {urls[i]} 失败: {result}") asyncio.run(main()) </code></pre> </div> With this fix, the <code>asyncio.sleep(0.5)</code> calls happen concurrently across all 10 tasks, compressing the total waiting time to roughly 0.5 seconds (plus the actual request time). Paired with aiohttp’s connection pooling, the efficiency gain is night and day. <h2> Lessons Learned: Ignore These and You’ll Trip Again </h2> <ol> <li>Check if your dependencies are truly async Simply sprinkling <code>async</code>/<code>await</code> into your code isn’t enough. Synchronous libraries like <code>time.sleep</code>, <code>requests</code>, or <code>pymongo</code> will immediately choke your event loop if used inside a coroutine. Always switch to their async equivalents: <code>aiohttp</code>, <code>httpx</code>, <code>motor</code> (async MongoDB driver), <code>aiomysql</code>, etc. If a library has no async version, offload it to a thread pool using <code>await asyncio.to_thread(sync_func, *args)</code>. It’s not perfect, but at least it won’t block the event loop.</li> <li>The <code>asyncio.gather</code> exception trap By default, <code>gather</code> will immediately propagate any exception thrown by a task, canceling other running tasks in the process. If you want all tasks to finish before handling results, remember to set <code>return_exceptions=True</code> and then manually inspect each returned value to see if it’s an <code>Exception</code> instance.</li> <li>Don’t abandon tasks created with <code>create_task</code> If you create a Task with <code>asyncio.create_task</code> but never <code>await</code> it or keep a reference, any exception it raises will be silently swallowed by garbage collection — you won’t even see an error log. Every Task should either be collected with <code>gather</code> or have its exception explicitly checked.</li> <li>Don’t spawn an unbounded number of coroutines Kicking off hundreds of concurrent connections when scraping hundreds of URLs can easily trip the target API’s rate limits. Always cap concurrency — for example, with an <code>asyncio.Semaphore</code> or by tuning your <code>aiohttp</code> connector pool — to avoid getting blocked or throttled.</li> </ol>

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。原文来自 — 版权归原作者所有。

推荐订阅源

DEV Community