Files
Eric Allam 60d71da90e perf(webapp,run-engine): cut CPU on the engine-facing worker-action routes (#4746)
Cuts CPU on the `engine/v1/worker-actions/*` routes a managed supervisor
calls, and adds the benchmark harness the numbers come from.

Measured on a local stack: **on-CPU per completed run 9.07ms → 6.59ms
(−27%)**, busy fraction 45.6% → 33.8%, with every worker-action p50 down
23–27%. Load was 5,000 runs / 24 virtual supervisors / 90s window /
30,120 requests / 0 errors.

Query-count work from the same investigation is deliberately **not**
here — it will follow as a separate PR.

## The three changes

**1. Split the event-loop monitor in two (~14% of on-CPU, plus ~5pp of
GC).**

`eventLoopMonitor.server.ts` installs a global `async_hooks` hook:
`init` writes a `Map` entry for *every* async resource the process
creates, `before` calls `process.hrtime()` and `context.active()` on
every one. Enabling any async hook also puts V8 on the slow path for
promise instrumentation process-wide. `EVENT_LOOP_MONITOR_ENABLED`
defaulted to `"1"`, so this was the shipping configuration.

The blocked-loop detector is now opt-in (`EVENT_LOOP_MONITOR_ENABLED`,
default `0`). The event-loop *utilization* gauge — a single interval
timer with no per-request cost — moves to its own flag
(`EVENT_LOOP_UTILIZATION_MONITOR_ENABLED`, default `1`) and stays on, so
the useful half survives without the expensive half.

A/B under identical load:

| | monitor on | monitor off | change |
|---|---|---|---|
| on-CPU per run | 9.08ms | 7.25ms | −20% |
| GC self time | 9.80% | 5.05% | −4.75pp |
| dequeue p50 | 76.6ms | 62.8ms | −18% |
| attempts/start p50 | 56.3ms | 43.5ms | −23% |

**2. Bucket route matching by first static path segment (10.4% → 3.9% of
on-CPU).**

`patches/@remix-run__router@1.23.3.patch` already memoized flattened
branches and compiled path regexes. What remained was the linear scan:
`matchRouteBranch` walked the ranked branch list calling `matchPath` per
branch across 521 route files, so every worker-action request paid a
scan proportional to the whole route table.

Branches are now indexed by their lowercased leading segment, with one
always-considered list for branches whose leading segment is dynamic,
splat or optional (and for root/pathless paths). A request walks only
its own bucket merged with that list. Route-matching self time dropped
64% (3.6s → 1.3s over a 90s window).

Ordering is preserved exactly: both lists hold indexes into the already
rank-sorted branch array and are walked in ascending-index order, so the
first match found is the same branch the full scan would have found.
Bucketing lowercases on both sides, so case-insensitive matching still
resolves and `caseSensitive: true` routes are still rejected by
`matchPath` itself. A pathname whose own leading segment can't be
bucketed falls back to the full scan.

Verified equivalent to the unpatched matcher over 20,050 pathnames
(literal, dynamic, splat, optional, case variants, basenames,
percent-encoded) with zero mismatches.
`apps/webapp/test/routeMatchingPatch.test.ts` pins the matching
semantics rather than the optimisation, so it still passes without the
patch.

**3. Demote per-heartbeat and per-dequeue `info` logs to `debug`.**

These are the two highest-rate engine calls and each wrote a synchronous
structured log line on every request. Synchronous `console` writes can
block the loop when stdout backs up, which costs more than the ~1.3% CPU
share suggests.

## The harness

Two benchmarks, neither in the default suite (they run for minutes,
attach the V8 profiler, and report numbers rather than assert on them).
See `apps/webapp/test/bench/README.md`.

- `apps/webapp/test/bench/engineHttp.bench.test.ts` — spawns a real
webapp against throwaway Postgres/Redis containers, seeds a production
environment with a promoted managed deployment, and drives a closed-loop
supervisor pool through the full lifecycle. Profiling runs over CDP
rather than `--cpu-prof` so it covers only the measured window instead
of being swamped by boot, and `performance.eventLoopUtilization()` is
sampled *inside* the webapp process.
-
`internal-packages/run-engine/src/engine/bench/runEngineLifecycle.bench.test.ts`
— drives `RunEngine` directly, profiling enqueue and lifecycle
separately so engine cost isn't mixed with request-stack overhead.
- `apps/webapp/test/bench/analyzeProfile.ts` — dependency-free
`.cpuprofile` analyzer that symbolicates through the build's source maps
and ranks CPU by package, self time and total time. Percentages are
shares of on-CPU time (V8's `(idle)`/`(program)` excluded).

`startWebapp` gains `overrideEnv`, applied after the worker-disable
defaults, so the HTTP bench can re-enable the run engine worker that
drains the master queue into the worker queues a supervisor dequeues
from.

The local OTel collector gains a traces pipeline. It only defined a
metrics pipeline, so pointing `INTERNAL_OTEL_TRACE_EXPORTER_URL` at it
locally failed and the webapp silently fell back to the console span
logger.

## Configuration

For operators upgrading:

- `EVENT_LOOP_MONITOR_ENABLED` (now defaults to `0`) — the
per-async-resource blocked-loop detector. Set to `1` to restore the
previous behaviour and keep emitting `event-loop-blocked` spans.
- `EVENT_LOOP_UTILIZATION_MONITOR_ENABLED` (new, defaults to `1`) — the
`nodejs.event_loop.utilization` gauge. Unchanged in behaviour; it just
has its own flag now so it survives turning the detector off.

## Notes for review

- `pnpm-lock.yaml` changes only because the router patch content
changed, which changes its patch hash.
- One thing the profile ruled out: with a real OTLP collector receiving
spans, tracing costs ~1.7% of on-CPU at 100% sampling and ~0.8% at the
production rate. Span shipping is not a hidden cost, so nothing here
touches it.
- Caveats on the numbers: a laptop, not production hardware, so DB and
Redis *latency* are unrepresentative (client-side CPU is what's ranked);
single webapp process; throughput varies ~5% run to run, which is why
the claims rest on on-CPU per run rather than req/s.

## Verification

- 20,050-pathname router equivalence check vs the unpatched matcher,
zero mismatches
- `apps/webapp/test/routeMatchingPatch.test.ts` (12 cases) passes
- webapp e2e smoke suite (68 tests) passes through the patched router
- run-engine suites covering the snapshot/attempt paths pass
- `typecheck`, `format`, `lint`, `knip` clean
2026-08-21 11:53:16 +01:00

253 lines
8.2 KiB
TypeScript

/**
* Minimal Chrome DevTools Protocol client for benchmarking a spawned webapp.
*
* The bench spawns the webapp with `--inspect=<port>` and drives the V8 CPU
* profiler over CDP rather than using `--cpu-prof`. Two reasons:
*
* 1. `--cpu-prof` only writes at process exit, so its profile covers boot,
* module init and shutdown as well as the load. Boot dominates a short run
* and buries the request-path frames this pass is about.
* 2. Over CDP the profiler can be started and stopped around the measured
* window only, and several separately-named profiles can be taken from a
* single webapp instance.
*
* The same connection samples `performance.eventLoopUtilization()` inside the
* target process, which is the number this pass is trying to move. Sampling it
* from the bench process would only describe the load generator.
*/
import { writeFile } from "node:fs/promises";
import { WebSocket } from "ws";
type CdpMessage = {
id?: number;
result?: unknown;
error?: { code: number; message: string };
};
export type EluSample = {
/** ms since the sampler started */
atMs: number;
/** utilization over the interval since the previous sample, 0..1 */
utilization: number;
};
export type EluStats = {
mean: number;
p50: number;
p95: number;
p99: number;
max: number;
sampleCount: number;
};
/**
* Node prints the inspector ws URL to stderr on boot, but the bench does not
* own the spawn, so discover it over the inspector's HTTP endpoint instead.
*/
async function discoverWebSocketUrl(port: number, timeoutMs = 30_000): Promise<string> {
const deadline = Date.now() + timeoutMs;
let lastError: unknown;
while (Date.now() < deadline) {
try {
const res = await fetch(`http://127.0.0.1:${port}/json/list`);
const targets = (await res.json()) as Array<{ webSocketDebuggerUrl?: string }>;
const url = targets.find((t) => t.webSocketDebuggerUrl)?.webSocketDebuggerUrl;
if (url) return url;
} catch (err) {
lastError = err;
}
await new Promise((r) => setTimeout(r, 200));
}
throw new Error(`No inspector target on port ${port} after ${timeoutMs}ms: ${lastError}`);
}
class CdpSession {
private ws: WebSocket;
private nextId = 1;
private pending = new Map<number, { resolve: (v: any) => void; reject: (e: Error) => void }>();
/**
* Anything that ends the socket has to settle the in-flight requests. If the
* profiled webapp exits mid-run, an unsettled `send()` would otherwise hang
* until the suite-level timeout with nothing explaining why.
*/
private constructor(ws: WebSocket) {
this.ws = ws;
this.ws.on("message", (data) => {
let msg: CdpMessage;
try {
msg = JSON.parse(data.toString()) as CdpMessage;
} catch {
return;
}
if (msg.id === undefined) return;
const waiter = this.pending.get(msg.id);
if (!waiter) return;
this.pending.delete(msg.id);
if (msg.error) waiter.reject(new Error(`${msg.error.message} (${msg.error.code})`));
else waiter.resolve(msg.result);
});
const rejectAll = (reason: string) => {
for (const waiter of this.pending.values()) {
waiter.reject(new Error(reason));
}
this.pending.clear();
};
this.ws.on("error", (err: Error) => rejectAll(`CDP socket error: ${err.message}`));
this.ws.on("close", () => rejectAll("CDP socket closed before the response arrived"));
}
/**
* A full CPU profile of a busy minute is tens of MB and arrives as a single
* ws frame, so the payload cap is raised well past the 100MB default.
*/
static async connect(inspectPort: number): Promise<CdpSession> {
const url = await discoverWebSocketUrl(inspectPort);
const ws = new WebSocket(url, { maxPayload: 512 * 1024 * 1024 });
await new Promise<void>((resolve, reject) => {
ws.once("open", () => resolve());
ws.once("error", reject);
});
return new CdpSession(ws);
}
send<T = any>(method: string, params: Record<string, unknown> = {}): Promise<T> {
if (this.ws.readyState !== WebSocket.OPEN) {
return Promise.reject(new Error(`CDP socket is not open, cannot send ${method}`));
}
const id = this.nextId++;
const promise = new Promise<T>((resolve, reject) => {
this.pending.set(id, { resolve, reject });
});
this.ws.send(JSON.stringify({ id, method, params }));
return promise;
}
close(): void {
this.ws.close();
}
}
export class WebappProfiler {
private session: CdpSession;
private eluTimer: NodeJS.Timeout | null = null;
private eluSamples: EluSample[] = [];
private eluStartedAt = 0;
private constructor(session: CdpSession) {
this.session = session;
}
static async attach(inspectPort: number): Promise<WebappProfiler> {
const session = await CdpSession.connect(inspectPort);
await session.send("Runtime.enable");
await session.send("Profiler.enable");
return new WebappProfiler(session);
}
/**
* `intervalUs` is V8's sampling interval in microseconds. The 200us default is
* 5x finer than V8's own 1ms: the engine routes are short, and at 1ms too few
* samples land inside a single request to separate the frames within it.
*/
async startCpuProfile(intervalUs = 200): Promise<void> {
await this.session.send("Profiler.setSamplingInterval", { interval: intervalUs });
await this.session.send("Profiler.start");
}
async stopCpuProfile(outPath: string): Promise<{ path: string; sampleCount: number }> {
const { profile } = await this.session.send<{ profile: { samples?: number[] } }>(
"Profiler.stop"
);
await writeFile(outPath, JSON.stringify(profile));
return { path: outPath, sampleCount: profile.samples?.length ?? 0 };
}
/**
* Awaits a baseline reading before the interval starts, so the first recorded
* delta is measured from the moment sampling started rather than from process
* boot. Awaiting matters: the baseline is a round trip to the target, and a
* tick that landed before it resolved would report a zero delta and drag the
* average down.
*/
async startEluSampling(intervalMs = 250): Promise<void> {
this.eluSamples = [];
this.eluStartedAt = Date.now();
await this.evaluateElu();
const timer = setInterval(() => {
void this.evaluateElu().then((utilization) => {
if (utilization !== undefined) {
this.eluSamples.push({ atMs: Date.now() - this.eluStartedAt, utilization });
}
});
}, intervalMs);
timer.unref();
this.eluTimer = timer;
}
/**
* Stashes the previous reading on globalThis inside the target so each call
* reports the delta since the last sample. A raw
* `performance.eventLoopUtilization()` is a since-boot average, which idle
* boot time drags down and which never recovers during a short run.
*/
private async evaluateElu(): Promise<number | undefined> {
try {
const res = await this.session.send<{ result: { value?: number } }>("Runtime.evaluate", {
expression: `(() => {
const now = performance.eventLoopUtilization();
const prev = globalThis.__benchLastElu;
globalThis.__benchLastElu = now;
if (!prev) return 0;
const diff = performance.eventLoopUtilization(now, prev);
return Number.isFinite(diff.utilization) ? diff.utilization : 0;
})()`,
returnByValue: true,
});
return res.result?.value;
} catch {
return undefined;
}
}
stopEluSampling(): { stats: EluStats; samples: EluSample[] } {
if (this.eluTimer) {
clearInterval(this.eluTimer);
this.eluTimer = null;
}
const samples = this.eluSamples;
if (samples.length === 0) {
return { stats: { mean: 0, p50: 0, p95: 0, p99: 0, max: 0, sampleCount: 0 }, samples };
}
const sorted = samples.map((s) => s.utilization).sort((a, b) => a - b);
const at = (q: number) => sorted[Math.min(sorted.length - 1, Math.floor(sorted.length * q))]!;
return {
stats: {
mean: sorted.reduce((a, b) => a + b, 0) / sorted.length,
p50: at(0.5),
p95: at(0.95),
p99: at(0.99),
max: sorted[sorted.length - 1]!,
sampleCount: sorted.length,
},
samples,
};
}
detach(): void {
this.stopEluSampling();
this.session.close();
}
}