8.6 KiB
8.6 KiB
id, title, category, status, canonical_id, aliases, duplicate_of, source_trust_level, confidence_score, verification_status, tags, raw_sources, last_reinforced, github_commit, tech_stack
| id | title | category | status | canonical_id | aliases | duplicate_of | source_trust_level | confidence_score | verification_status | tags | raw_sources | last_reinforced | github_commit | tech_stack | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| wiki-2026-0508-chrome-devtools-memory-profiling | Chrome DevTools Memory Profiling | 10_Wiki/Topics | verified | self |
|
none | A | 0.93 | applied |
|
2026-05-10 | pending |
|
Chrome DevTools Memory Profiling
📌 한 줄 통찰
"매 heap 의 snapshot 의 leak 의 detect". 매 V8 GC 의 못 하는 reference 의 trace. 매 retaining path 의 root cause. 매 3-snapshot 기법 + 매 allocation timeline 의 modern leak hunt 의 standard.
📖 핵심
매 3 tool
1. Heap Snapshot
- 매 specific 시점 의 entire object graph.
- 매 GC 가 매 force 후 의 capture.
- 매 view: Summary, Comparison, Containment.
2. Allocation Timeline (Allocation instrumentation)
- 매 매 50ms 의 allocation track.
- 매 stack trace 의 keep.
- 매 blue bar = 매 alive, 매 grey = 매 GC'd.
3. Allocation Sampling
- 매 statistical sampling.
- 매 production-friendly (low overhead).
매 3-snapshot 기법
- Snapshot 1: 매 baseline (cleanup 의 후).
- 매 suspect action 의 N times.
- Snapshot 2: 매 mid.
- 매 same action 의 N times.
- Snapshot 3: 매 final.
- 매 Comparison 의 #2 → #3.
- 매 새 object 의 leak candidate.
매 leak source (common)
Detached DOM
- 매 element 의 remove 가 매 reference 가 keep.
Closure trap
- 매 function 의 outer scope 의 large object 의 capture.
Forgotten timer / listener
- 매 setInterval / addEventListener 의 cleanup 의 X.
Console.log
- 매 logged object 의 reference 의 hold (devtools open 시).
WebSocket / fetch retain
- 매 connection 의 close 의 X.
Cache unbounded
- 매 Map / Set 의 grow 의 limit X.
매 retaining path (Retainers)
- 매 GC root → 매 object 의 chain.
- 매 root: 매 window, document, persistent store.
- 매 chain 의 break = 매 fix.
매 GC type (V8)
- Scavenge (young gen): 매 frequent, fast.
- Mark-Compact (old gen): 매 less frequent, expensive.
- Incremental: 매 spread over frame.
Heap profile vs Performance
| Tool | 측정 |
|---|---|
| Memory tab | 매 heap snapshot, allocation, leak |
| Performance tab | 매 timing + 매 memory chart |
| Coverage tab | 매 unused JS / CSS |
| Lighthouse | 매 budget audit |
💻 패턴
Code (intentional leak — for testing)
// 매 detached DOM leak
let leaked = [];
function leak() {
const el = document.createElement('div');
el.innerHTML = 'x'.repeat(1_000_000);
document.body.appendChild(el);
document.body.removeChild(el);
leaked.push(el); // 매 reference 의 keep — 매 leak
}
Profiling steps (DevTools)
1. Memory tab → Heap snapshot → Take snapshot (snapshot 1).
2. 매 suspect action 의 5-10 times.
3. Take snapshot (snapshot 2).
4. 매 same action 의 5-10 times.
5. Take snapshot (snapshot 3).
6. View dropdown → "Comparison" between 2 and 3.
7. Sort by "# Delta" or "Size Delta" descending.
8. Look for:
- Detached HTMLDivElement (or similar)
- Closure
- (compiled code) holding objects
9. Click constructor → object → "Retainers" panel.
10. Trace path to GC root.
Common fix patterns
// 매 ❌ Leak
class Component {
constructor() {
this.handler = () => this.doSomething();
window.addEventListener('resize', this.handler);
setInterval(this.handler, 1000);
}
destroy() {
// 매 forgot to remove
}
}
// 매 ✅ Fix
class Component {
constructor() {
this.handler = () => this.doSomething();
this.intervalId = setInterval(this.handler, 1000);
window.addEventListener('resize', this.handler);
}
destroy() {
window.removeEventListener('resize', this.handler);
clearInterval(this.intervalId);
this.handler = null;
}
}
WeakRef / WeakMap (modern)
// 매 cache 의 강제 keep 의 X
const cache = new WeakMap();
function getMetadata(obj) {
if (!cache.has(obj)) cache.set(obj, computeExpensive(obj));
return cache.get(obj);
}
// 매 obj 의 GC 시 의 entry 도 sweep.
performance.memory API
// 매 Chromium 만 (deprecated 가, 매 still useful)
function logMemory(label) {
if (performance.memory) {
const m = performance.memory;
console.log(`${label}: used=${(m.usedJSHeapSize / 1048576).toFixed(1)}MB, ` +
`total=${(m.totalJSHeapSize / 1048576).toFixed(1)}MB`);
}
}
logMemory('before');
heavyOperation();
logMemory('after');
measureUserAgentSpecificMemory (modern, secure)
async function checkMemory() {
if (!('measureUserAgentSpecificMemory' in performance)) return;
const result = await performance.measureUserAgentSpecificMemory();
console.log(`Total: ${(result.bytes / 1048576).toFixed(1)} MB`);
for (const breakdown of result.breakdown) {
console.log(` ${breakdown.types.join('+')}: ${(breakdown.bytes / 1048576).toFixed(1)} MB`);
}
}
→ 매 cross-origin isolated 환경 만.
Detect leaked listener (test)
function getEventListenerCount(element) {
// 매 Chrome devtools console 만
return getEventListeners(element);
}
// 매 test
const before = getEventListenerCount(window);
mountAndUnmountComponent();
const after = getEventListenerCount(window);
if (Object.keys(after).length > Object.keys(before).length) {
console.error('Listener leak detected');
}
React-specific (heap on unmount)
import { useEffect } from 'react';
function Comp() {
useEffect(() => {
const handler = () => {};
window.addEventListener('scroll', handler);
const id = setInterval(() => {}, 1000);
// 매 cleanup
return () => {
window.removeEventListener('scroll', handler);
clearInterval(id);
};
}, []);
return <div />;
}
Chrome Lighthouse memory budget
# 매 lighthouse 의 audit (CI)
- name: Lighthouse memory check
uses: treosh/lighthouse-ci-action@v10
with:
urls: 'https://staging.example.com/'
budgetPath: '.lighthouse/budget.json'
[{
"resourceSizes": [
{ "resourceType": "script", "budget": 300 },
{ "resourceType": "total", "budget": 1000 }
]
}]
Headless leak detection (Puppeteer)
const puppeteer = require('puppeteer');
async function detectLeak() {
const browser = await puppeteer.launch();
const page = await browser.newPage();
await page.goto('https://example.com');
const before = (await page.metrics()).JSHeapUsedSize;
for (let i = 0; i < 100; i++) {
await page.click('#leak-button');
}
// 매 GC force
await page.evaluate(() => globalThis.gc?.()); // requires --js-flags="--expose-gc"
const after = (await page.metrics()).JSHeapUsedSize;
if (after / before > 2) {
console.error(`Memory leak: ${before} → ${after}`);
}
await browser.close();
}
🤔 결정 기준
| 상황 | Tool |
|---|---|
| One-off leak hunt | Heap snapshot 3-shot |
| Allocation pattern | Allocation Timeline |
| Production prof | Allocation Sampling |
| Continuous (CI) | Puppeteer + metrics |
| Cross-origin | measureUserAgentSpecificMemory |
| React | DevTools Profiler + Memory |
기본값: 매 heap snapshot 3-shot + Comparison + Retainer trace.
🔗 Graph
- 부모: Web-Performance · Browser · Memory-Management
- 변형: Heap-Snapshot · Allocation-Timeline · Garbage-Collection · V8
- 응용: Memory-Leak-Detection · WeakMap · WeakRef · Lighthouse
- Adjacent: Bottlenecks · CSS Animations · Frontend-Performance
🤖 LLM 활용
언제: 매 memory leak hunt. 매 SPA performance audit. 매 long-running tab. 언제 X: 매 server-side. 매 CPU bottleneck (Performance tab).
❌ 안티패턴
- No GC force: 매 stale snapshot.
- No baseline (1-shot): 매 noisy.
- Console.log 의 evidence 의 ignore: 매 false leak.
- Production 의 Allocation Timeline: 매 overhead.
- Detached DOM 무시: 매 typical 의 root cause.
- Listener cleanup X: 매 most leak.
🧪 검증 / 중복
- Verified (Chrome DevTools docs, V8 blog).
- 신뢰도 A.
- Related: Web-Performance · Bottlenecks · CSS Animations · Baseline-Project.
🕓 Changelog
| 날짜 | 변경 |
|---|---|
| 2026-04-19 | Auto-mapped |
| 2026-05-08 | Phase 1 |
| 2026-05-10 | Manual cleanup — 3 tool + 3-snapshot + leak source + 매 Puppeteer / WeakRef code |