惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

V
V2EX - 技术
D
DataBreaches.Net
阮一峰的网络日志
阮一峰的网络日志
Recent Announcements
Recent Announcements
V
V2EX
Hugging Face - Blog
Hugging Face - Blog
T
The Exploit Database - CXSecurity.com
Simon Willison's Weblog
Simon Willison's Weblog
Cisco Talos Blog
Cisco Talos Blog
Microsoft Security Blog
Microsoft Security Blog
C
Cyber Attacks, Cyber Crime and Cyber Security
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
K
Kaspersky official blog
F
Fortinet All Blogs
GbyAI
GbyAI
Forbes - Security
Forbes - Security
The Cloudflare Blog
博客园 - 司徒正美
博客园_首页
量子位
Schneier on Security
Schneier on Security
G
GRAHAM CLULEY
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
P
Proofpoint News Feed
N
News | PayPal Newsroom
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
博客园 - 聂微东
T
Tor Project blog
V
Vulnerabilities – Threatpost
Y
Y Combinator Blog
Jina AI
Jina AI
Help Net Security
Help Net Security
T
Threat Research - Cisco Blogs
Recent Commits to openclaw:main
Recent Commits to openclaw:main
C
Cybersecurity and Infrastructure Security Agency CISA
Project Zero
Project Zero
N
News and Events Feed by Topic
I
Intezer
B
Blog
美团技术团队
C
CERT Recently Published Vulnerability Notes
NISL@THU
NISL@THU
L
LINUX DO - 最新话题
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
Blog — PlanetScale
Blog — PlanetScale
AWS News Blog
AWS News Blog
T
Tailwind CSS Blog
The Last Watchdog
The Last Watchdog
雷峰网
雷峰网
有赞技术团队
有赞技术团队

The Practical Developer

The Libuv Thread Pool Trap: Why Node.js Async APIs Stall Under Load Postgres Covering Indexes with INCLUDE: Eliminate Heap Fetches on Read-Heavy Workloads Postgres DISTINCT ON: The Fastest Way to Get the Latest Row Per Group Postgres Transaction Isolation: The Anomalies Your App Actually Faces in Production Linux TCP Tuning for Node.js Microservices: The Kernel Settings That Stop Silent Connection Drops Under Load Postgres HOT Updates and Fillfactor: Why Not All Writes Are Created Equal Database Connection Pool Leaks: Finding the Promise That Never Returns Its Seat Linux OOM Killer in Production: Why Your Node.js Containers Die Without a Stack Trace Postgres Materialized Views: Refresh Strategies That Do Not Lock Your Dashboards API Dependency Health Checks: Why /health Is Not Enough Authorization with Zanzibar Tuples: How Google Manages Permissions and How To Build the Same Check in Node.js Postgres Advisory Locks: The 20-Character Primitive That Replaces Redis for Coordination Dead Letter Queues: The Message Queue Pattern That Saves You at 2 a.m. File Descriptor Exhaustion: The Kernel Limit That Silently Drops Node.js Connections Graceful Degradation: The Pattern That Turns Total Outages into Partial Success PostgreSQL Full-Text Search: Dropping Elasticsearch for 90% of Use Cases S3 Presigned Multipart Uploads: Stop Your API Server from Being a File Upload Bottleneck MessagePack vs JSON: The Binary Serialization Switch That Cut Our Internal RPC Overhead by 40% DNS Caching in Node.js: The Silent Cause of Production Latency Spikes Reliable Cron Jobs: The Pattern That Stops Double Runs, Missed Executions, And The 2 AM Page GraphQL Query Complexity: Stop the OOM Query Before It Reaches Your Resolver Node.js Event Loop Lag: The Hidden Metric Behind Random Latency Spikes API Request Validation with Zod: The Schema That Catches Bad Input Before It Corrupts Your Database Load Shedding in Node.js: How to Reject Traffic Before You Drown Request Hedging: Cut Tail Latency In Half Without Overprovisioning Git Bisect: The Automated Binary Search That Finds Breaking Commits in Minutes Node.js Garbage Collection Tuning: Stop Letting V8 Pause Your Event Loop Node.js Server Timeouts: The Settings That Stop Slow Clients from Holding Sockets Hostage Postgres BRIN Indexes: The Time-Series Secret That Shrinks Indexes by 99% Event Sourcing with PostgreSQL: The Pragmatic 80% Solution Node.js Cluster Mode: Scaling the Event Loop Across CPU Cores Postgres Partial Indexes: Stopping Soft Deletes from Ruining Your Query Performance Request Coalescing with the Singleflight Pattern: Stop Drowning Your Database on Every Cache Miss The Bulkhead Pattern: Why One Slow Endpoint Should Not Drown Your Whole Service Node.js AsyncLocalStorage: End-to-End Request Context Without the Propagation Hell Postgres Deadlocks: Logging the Victim, Reproducing the Race, and Fixing the Lock Order Your Node.js HTTP Client Is the Bottleneck: Connection Pool Tuning That Works Optimistic Locking in Postgres: Stop Losing Data to Race Conditions Postgres Read Replicas: Stop Serving Stale Data to Your Users Cursor Pagination: Why Offset Queries Explode at Scale and How to Fix Them Node.js Worker Threads: 60 Lines That Stop a CSV Upload from Timing Out Every Other Request Reliable Webhook Delivery: Architecture for Outbound HTTP You Can Trust Request Timeouts and Deadline Propagation: Stop the Chain of Slowness Advanced Security Practices in Node.js Graceful Shutdown in Node.js: The 40 Lines That Stop 502s During Deploys Finding Node.js Memory Leaks with Heap Snapshots Idempotency Keys in 30 Lines: Stop Your Webhook From Charging Customers Twice Backpressure In Node.js: The Fix For Slow-Motion Queue Meltdowns Retries Done Right: Jitter, Budgets, and the Stampede You Did Not See Coming The Cache Stampede: Why Your "Just Add Redis" Layer Crashes Postgres at 3 a.m. Postgres SKIP LOCKED: An 80-Line Job Queue You Can Run Without Redis Stop Doing Work Nobody Wants: AbortController in Node.js, Done Right The N+1 Query Problem: We Found 23 In One Codebase And Killed Every One I Tried 5 AI Coding Tools for a Month. Here Is What I Actually Use CI/CD From Zero to Production in 30 Minutes With GitHub Actions Node.js vs Bun vs Deno: Which Runtime Should You Pick in 2025? Kubernetes Resource Requests And Limits: The Numbers That Decide If Your Cluster Is Stable The Three Pillars of Observability Are A Myth: What Actually Matters In Production pnpm Vs npm Vs yarn Vs Bun For Monorepos: Which One Earns The Migration In 2024 JSONB Indexing In Postgres: GIN Vs Expression Indexes, And When Each Is The Right Choice A Code Review Checklist That Ends The Same Three Arguments Every Sprint gRPC Vs REST In 2024: When The Switch Pays For Itself React Suspense For Data Fetching: The Pattern That Replaces Half Your Loading State Code The Five-Stage Rollout: How To Ship A Risky Change Without Holding Your Breath GitHub Actions In A Monorepo: Caching, Path Filters, And Secret Boundaries That Actually Work The Blameless Postmortem That Actually Improves Things: A Template And Six Hard-Won Rules Recursive CTEs In Postgres: How To Query A Tree Without N Round Trips Node.js Streams: When They Actually Help, And When They Just Add Complexity Playwright Vs Cypress In 2024: The Honest Comparison Of Which One Earns The Test Time Pod Disruption Budgets: The K8s Object That Keeps Your Service Up During Cluster Maintenance Postgres LISTEN/NOTIFY: The Pub/Sub You Already Have And Are Not Using Chaos Engineering Starter Kit: The Five Drills That Don't Need Netflix-Scale Spec-Driven API Development With OpenAPI: How To Stop Drifting From Your Docs Saga Pattern vs Two-Phase Commit: Distributed Transactions Without The Lies Kubernetes Autoscaling Beyond CPU: The Custom-Metric HPA Pattern That Actually Works Postgres Partitioning For Time-Series: The Boring Setup That Saves Your Database Distributed Locks With Redis: An Honest Look At Redlock And When You Don't Need It HTTP/2 vs HTTP/3: What Actually Changes For Your App, And What Doesn't Image Optimization For The Web In 2023: srcset, AVIF, And The Lighthouse Score You Actually Want Kafka vs RabbitMQ: A Decision Tree That Doesn't Hate You UUID vs Bigint Primary Keys In Postgres: The Index Math That Decides For You Flame Graphs: How To Find The Slow Function In 30 Seconds Without Profiling Theatre Postgres Streaming Vs. Logical Replication: Which One Solves Your Actual Problem ESLint Rules That Earn Their Keep: The Twelve I Enable On Every Project Pre-Commit Hooks That Pay For Themselves: Husky, lint-staged, And The Five Rules That Stick Zero-Downtime Database Migrations: The Six-Step Pattern That Rules Them All Circuit Breakers In Node.js: 50 Lines That Stop A Failing Dependency From Taking Down Your Service Postgres VACUUM Is Not Magic: How Your Hot Table Bloats To 80GB And How To Fix It Kubernetes Liveness And Readiness Probes: The Difference That Causes Half Your Outages Rate Limiting In Production: A Token Bucket In 30 Lines Of Redis The Outbox Pattern: How To Stop Losing Events When Postgres And Kafka Disagree Load Testing With k6: The Three Scenarios That Find Real Bugs (Not Synthetic Numbers) Postgres Row-Level Security For Multi-Tenant Apps: The Pattern That Stops You From Leaking Data Rebase vs. Merge: The Team Policy That Ends The Argument Forever OpenTelemetry in Node.js: Distributed Tracing That Actually Helps During an Incident Feature Flags That Pay Rent: The 4 Flag Types And When To Delete Each ETag, Last-Modified, and the Caching Headers Most APIs Get Wrong Connection Pooling Without the Cargo Cult: pgbouncer in 100 Lines of Config JSONB Is Not a Schema: When To Reach For It in Postgres, And When To Stop Bash Strict Mode: The Three Lines That Stop Your Deploy Script From Lying To You
React Server Components: The Mental Model That Makes The "use client" Boundary Obvious
The Practica · 2024-02-02 · via The Practical Developer

The team starts a Next.js 14 project. The first PR is full of "use client" at the top of every component, “just to be safe.” A senior engineer asks why. Nobody knows; everyone is confused about which components should be server and which client. The bundle size for the homepage is 800 KB of JS for a page that mostly displays static content.

React Server Components (RSC) is the most-misunderstood frontend feature of the past few years. The reason is mostly that the mental model from “regular React” doesn’t translate. Once you have the right model, the "use client" boundary becomes an obvious answer to one question: does this component need to be reactive on the user’s machine?

This post is the model, applied to a few real component patterns, with the four traps that cause most teams to mis-mark components.

The core idea

A React Server Component:

  • Renders only on the server.
  • Has access to the database, file system, secrets, and any Node.js API.
  • Ships its output as HTML/JSON to the browser, but the component code itself is not in the bundle.
  • Cannot use hooks (useState, useEffect), event handlers (onClick), or browser APIs.

A Client Component:

  • Renders on both server (SSR) and client.
  • Hydrates on the client; can use hooks, events, browser APIs.
  • The code is in the bundle.

The boundary between them is "use client" at the top of a file. Anything below that file’s exports is a client component (and so are any imports).

The mental model: reactivity boundary

Ask one question per component: “Does this component need to be reactive on the user’s machine?”

  • “It just shows data” → server component.
  • “It has a button that does something” → client component.
  • “It uses useState, useEffect, etc.” → client component.
  • “It needs to access the URL or browser APIs” → client component.

Most components (typography, layout, anything that just renders props) should be server components. The “client” boundary should be small islands at the leaves of the tree, not the whole tree.

A worked example

A typical product page:

// app/products/[id]/page.tsx - Server Component (default in Next.js app dir)
import { db } from '@/lib/db';
import { AddToCartButton } from './add-to-cart-button';
import { Reviews } from './reviews';

export default async function ProductPage({ params }: { params: { id: string } }) {
  const product = await db.products.findUnique({ where: { id: params.id } });

  return (
    <div>
      <h1>{product.name}</h1>
      <p>{product.description}</p>
      <p>${product.priceCents / 100}</p>
      <AddToCartButton productId={product.id} />
      <Reviews productId={product.id} />
    </div>
  );
}
// app/products/[id]/add-to-cart-button.tsx - Client Component
'use client';
import { useState } from 'react';

export function AddToCartButton({ productId }: { productId: string }) {
  const [adding, setAdding] = useState(false);

  return (
    <button
      onClick={async () => {
        setAdding(true);
        await fetch(`/api/cart`, { method: 'POST', body: JSON.stringify({ productId }) });
        setAdding(false);
      }}
      disabled={adding}
    >
      {adding ? 'Adding...' : 'Add to cart'}
    </button>
  );
}
// app/products/[id]/reviews.tsx - Server Component
import { db } from '@/lib/db';

export async function Reviews({ productId }: { productId: string }) {
  const reviews = await db.reviews.findMany({ where: { productId } });
  return (
    <ul>
      {reviews.map(r => <li key={r.id}>{r.rating} - {r.body}</li>)}
    </ul>
  );
}

Notice: the page accesses the database directly. The reviews component does too, server-to-server, no API call needed. The Add-to-Cart button is the only client component, and only because it has a useState and a click handler. Bundle size for this page: just the button.

The four traps

1. Marking a parent as "use client" because a child needs it. A common reaction: “the button needs "use client", so let me mark the page too.” Wrong direction. The "use client" directive applies to the file where it appears and any imports below it. The parent stays a server component; only the leaf component is client.

2. Trying to use hooks in a server component. A server component has no useState, useEffect, or useRef. If you need them, the component must be client. The error message (“Hooks can only be used inside the body of a function component”) points at this.

3. Passing functions from server to client components. Server components can pass props to client components, but functions are not serializable.

// SERVER COMPONENT:
const onSubmit = async (data) => { /* ... */ };
return <ClientForm onSubmit={onSubmit} />;  // ← error: function is not serializable

The fix: use Server Actions (async function action(...) marked "use server").

4. Importing client-only libraries in server components. A library that uses window, document, or browser APIs cannot run on the server. Either: "use client" the importing component, or wrap the library in a dynamic import that runs only on the client.

Server Actions

Server Actions are the missing piece that makes RSC ergonomic for forms and mutations:

// app/actions.ts
'use server';

import { db } from '@/lib/db';

export async function addToCart(productId: string) {
  await db.cart.upsert({
    where: { userId: getCurrentUserId() },
    create: { items: [{ productId }] },
    update: { items: { push: { productId } } },
  });
}
// AddToCartButton.tsx - Client Component
'use client';
import { addToCart } from '@/app/actions';

export function AddToCartButton({ productId }: { productId: string }) {
  return (
    <form action={() => addToCart(productId)}>
      <button>Add to cart</button>
    </form>
  );
}

The button is a client component (event handler, possibly useState for loading); the action runs on the server and has direct DB access. No API endpoint to write.

The right boundary placement

A heuristic that works: draw the smallest possible client component, push it as far down the tree as possible.

If you have a large component that mostly renders data and has one interactive piece, refactor:

// Before: whole component is client because of one button.
'use client';
export function ArticlePage({ articleId }) {
  const [bookmarked, setBookmarked] = useState(false);
  // ... 200 lines of mostly static rendering ...
  return <div>{/* ... */} <button onClick={...}>bookmark</button></div>;
}

// After: page is server, button is a tiny client component.
export async function ArticlePage({ articleId }) {
  const article = await db.articles.findUnique(...);
  return <div>
    <h1>{article.title}</h1>
    <p>{article.body}</p>
    <BookmarkButton articleId={article.id} />
  </div>;
}

Bundle size collapses. Most rendering happens on the server.

Streaming and Suspense

RSC integrates with React Suspense for streaming:

import { Suspense } from 'react';

export default async function Page() {
  return (
    <div>
      <Header />  {/* renders fast */}
      <Suspense fallback={<Spinner />}>
        <SlowReviews productId="..." />  {/* streamed when ready */}
      </Suspense>
    </div>
  );
}

The browser receives the header HTML immediately and streams the Reviews HTML when the database query finishes. Time-to-first-byte stays fast, time-to-interactive is correct.

This is one of RSC’s bigger wins: it makes streaming the default rather than an opt-in optimization.

Caching and data fetching

In Next.js App Router, fetch() is automatically cached and deduplicated across components in the same request:

// Both components fetch, but Next.js makes only one network call.
async function ParentComp() {
  const data = await fetch('https://api.example.com/foo').then(r => r.json());
  // ...
}

async function ChildComp() {
  const data = await fetch('https://api.example.com/foo').then(r => r.json());
  // ...
}

Customize via { next: { revalidate: 60 } } (cache for 60 seconds) or { cache: 'no-store' } (always fresh).

For database calls, you implement caching yourself or use cache() from React.

When NOT to use RSC

A few cases:

  • Mostly-interactive apps (a SaaS dashboard with rich UI everywhere). Pure SPA might be cleaner.
  • Heavy reliance on a non-RSC-ready ecosystem. Some libraries don’t work as server components yet; check before committing.
  • Edge runtime constraints. RSC works best with full Node.js; some edge runtimes have limitations.

For most marketing sites, content sites, and admin panels with mixed interactivity, RSC is the right default.

The takeaway

The mental model: “does this component need to be reactive on the user’s machine?” If yes, client component. If no, server component. Most components are server components; client components are small leaves at the bottom of the tree. Pass data down, mark interactive bits as client, use Server Actions for mutations.

The team that internalizes this ships pages with 50 KB of JS instead of 500 KB. The ones that don’t paste "use client" at the top of everything and lose the entire benefit.


A note from Yojji

The kind of frontend architecture judgment that takes advantage of new React capabilities, keeping bundles small with the right server/client split, is the kind of detail Yojji’s frontend teams put into the products they ship for clients.

Yojji is an international custom software development company founded in 2016, with teams across Europe, the US, and the UK. They specialize in the JavaScript ecosystem (React, Node.js, TypeScript), cloud platforms, and full-cycle product engineering, including the React Server Component architecture decisions that decide whether a site feels fast or merely modern.