Measured on production after the 4.5.0 deploy: with a new connection per request, 1,000 requests in 53 s got no 429 from RL_SITE (120/60 s); over one keep-alive connection, 429 after ~93. The ratelimits binding counts per machine of a data centre and syncs late, so the zone's WAF rule is the limit that holds against floods. Verified once the owner set the rule (300/10 s): 800 requests in 4.9 s gave 111 x 200, 201 x Worker 429 and 488 x WAF 1015, with the site answering again 10 s later. Runbook test commands, code comments, CLAUDE.md gotcha 43, the 4.5.0 changelog entry. [skip-discuss] Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
65 lines
3.2 KiB
TypeScript
65 lines
3.2 KiB
TypeScript
/**
|
|
* Per-client request limits for everything that runs the Worker: the second
|
|
* layer behind the zone's WAF rate-limiting rule (docs/operations.md §1
|
|
* step 15). Static assets (/assets/*, /data/pack/*, …) are served without
|
|
* running the Worker, so only the WAF rule limits those — and only the WAF
|
|
* saves the Worker request a flood would cost.
|
|
*
|
|
* One `ratelimits` binding per kind of traffic (wrangler.jsonc), keyed by the
|
|
* client's IPv4 address or IPv6 /64 (rateKey):
|
|
*
|
|
* RL_IMG /img/* covers (a Games page shows up to 100)
|
|
* RL_ADMIN /admin, /admin/*, /api/admin/* the maintainer (Access checks who first)
|
|
* RL_SITE the rest: app pages, the public /api, the readable /data JSON, misses
|
|
*
|
|
* Suggest and the gate's POST keep their own, stricter limits on top
|
|
* (RL_SUGGEST, RL_GATE). Verified bots (the gate's Transform Rule header) are
|
|
* not counted, so a crawl never rate-limits itself out of the index. Counts
|
|
* are approximate (the binding's design): each machine of a data centre
|
|
* counts on its own and syncs late, so a client that keeps one connection
|
|
* (a browser, a script with keep-alive) is held to the limit, but one that
|
|
* opens a new connection per request is barely counted — the WAF rule is the
|
|
* limit for floods. Fails open: a missing binding or a failing limiter lets
|
|
* the request through.
|
|
*/
|
|
import type { WorkerEnv } from './env'
|
|
import { botPass } from './gate'
|
|
import { rateLimitedPage } from './gate-page'
|
|
import { json, rateKey } from './http'
|
|
|
|
/** `period` of every limit in wrangler.jsonc (ratelimit.test.ts checks), sent as Retry-After. */
|
|
export const RATE_PERIOD = 60
|
|
|
|
export type RateBucket = 'RL_SITE' | 'RL_IMG' | 'RL_ADMIN'
|
|
|
|
export function rateBucket(path: string): RateBucket {
|
|
if (path.startsWith('/img/')) return 'RL_IMG'
|
|
if (path === '/admin' || path.startsWith('/admin/') || path.startsWith('/api/admin/')) return 'RL_ADMIN'
|
|
return 'RL_SITE'
|
|
}
|
|
|
|
/** The 429: JSON for the API, covers, data and files; the "too many requests" page for app pages. */
|
|
export function tooManyRequests(request: Request, url: URL): Response {
|
|
const path = url.pathname
|
|
const retry = { 'Retry-After': String(RATE_PERIOD) }
|
|
// CORS on /data, so the desktop/Android apps see the 429, not a CORS error.
|
|
if (path.startsWith('/data/')) return json({ error: 'rate_limited' }, 429, { ...retry, 'Access-Control-Allow-Origin': '*' })
|
|
const last = path.slice(path.lastIndexOf('/') + 1)
|
|
if (/^\/(api|img|assets)\//.test(path) || last.includes('.')) return json({ error: 'rate_limited' }, 429, retry)
|
|
return rateLimitedPage(request, RATE_PERIOD)
|
|
}
|
|
|
|
/** Counts the request; null lets it through, else the 429 to answer with. */
|
|
export async function rateLimit(request: Request, url: URL, env: WorkerEnv): Promise<Response | null> {
|
|
const limiter: Pick<RateLimit, 'limit'> | undefined = env[rateBucket(url.pathname)]
|
|
if (!limiter) return null
|
|
if (await botPass(request, env.GATE_BOT_TOKEN ?? '')) return null
|
|
try {
|
|
if ((await limiter.limit({ key: rateKey(request) })).success) return null
|
|
} catch (e) {
|
|
console.error('rate limit: limiter failed, request let through', e)
|
|
return null
|
|
}
|
|
return tooManyRequests(request, url)
|
|
}
|