mirror of
https://github.com/MBombeck/HealthLog.git
synced 2026-11-07 04:12:04 +00:00
perf(bench): add a large-account seed and a request benchmark
scripts/bench/seed-large-account.ts (moved from scripts/, extended) writes one synthetic account of ten years: 1,265,561 measurement rows across the main types and five sources (Apple Health dense spot rows, Withings blood pressure and scale, a weekly manual weight, three years of Oura nightly values, 90 days of Nightscout CGM), `stats:` and per-sample cumulative rows, nightly sleep stages, 3,020 workouts, 1,095 mood entries and three medications with 3,285 intake events. Everything goes in through set-based INSERT … SELECT generate_series; values and timestamps are functions of the row index and `--anchor`, so a run is reproducible. The rollup tier is folded with recomputeUserRollups, as the boot backfill does, and the tables are analysed. About 80 s on the Mac mini. scripts/bench/run.ts boots the production build once with every role (so the queues exist), then measures against a web-only process under the production heap ceiling: p50, p95 and max latency and SQL statements per request (pg_stat_statements, read once nothing is in flight and the count holds still) for /api/auth/me, the dashboard summary and snapshot, the pulse series over 30, 365 and 3650 days, the day view, insights health status and correlations, each cold (caches evicted by a one-row manual-sync ingest before every request) and warm, plus a 5,000-row batch ingest as ten requests of 500. It writes JSON and a markdown table. docs/contributing/benchmarks.md describes the setup against a throwaway Postgres in Docker with the production pg_stat_statements settings.
This commit is contained in:
Marc-André Bombeck committed
1 parent
45ea61261e
commit
375ca12b05
7 files changed
+1225
-297
No files matched your search
@@ -59,6 +59,9 @@ notes, and machine-readable specs that ship alongside the source:
|
||||
- [`contributing/ui-standards.md`](./contributing/ui-standards.md) — the
|
||||
binding UI standards: card anatomy, spacing, text colour, tokens and
|
||||
which primitive to use where. Read before touching any surface.
|
||||
- [`contributing/benchmarks.md`](./contributing/benchmarks.md) — the
|
||||
large-account benchmark (latency and statements per request) and the
|
||||
statement budgets on the hot routes.
|
||||
- [`ui-guidelines.md`](./ui-guidelines.md) — UI / a11y house style; where
|
||||
it disagrees with the UI standards, the standards win.
|
||||
|
||||
|
||||
@@ -0,0 +1,83 @@
|
||||
# Benchmarks
|
||||
|
||||
Two tools measure the server's read and write paths. Use them before and after any change to a hot path, and put both numbers in the PR.
|
||||
|
||||
- **`scripts/bench/run.ts`** measures latency (p50, p95, max) and SQL statements per request against a seeded account of about 1.26 million measurement rows, through the production build. It is a local tool, not a CI gate: latency depends on the machine.
|
||||
- **`tests/integration/query-budget-hot-routes.test.ts`** holds a statement budget per hot route on a small fixture. Statement counts are exact on a fixed fixture, so this one is a gate: a change that adds a query per row, or a second read of something already loaded, fails it.
|
||||
|
||||
## The large-account benchmark
|
||||
|
||||
### 1. A throwaway Postgres with pg_stat_statements
|
||||
|
||||
Never point the seed or the benchmark at a database that holds real data. They write under a synthetic user, and the seed's `--reset` deletes that user with everything it owns.
|
||||
|
||||
```sh
|
||||
docker run -d --name hl-bench-db --shm-size=256m \
|
||||
-e POSTGRES_USER=bench -e POSTGRES_PASSWORD=bench -e POSTGRES_DB=healthlog_bench \
|
||||
-p 55432:5432 postgres:16.15-alpine \
|
||||
postgres -c shared_preload_libraries=pg_stat_statements -c pg_stat_statements.track=top
|
||||
```
|
||||
|
||||
The image and the two settings are the ones `docker-compose.yml` runs in production.
|
||||
|
||||
### 2. Environment, migrations, seed
|
||||
|
||||
```sh
|
||||
export DATABASE_URL="postgresql://bench:bench@127.0.0.1:55432/healthlog_bench?schema=public"
|
||||
export ENCRYPTION_KEYS='{"v1":"<64 hex>"}' ENCRYPTION_ACTIVE_KEY_ID=v1
|
||||
export API_TOKEN_HMAC_KEY="<64 hex>"
|
||||
|
||||
pnpm db:migrate:deploy
|
||||
pnpm dlx tsx scripts/bench/seed-large-account.ts --reset --anchor 2026-10-22
|
||||
```
|
||||
|
||||
Any throwaway key values work, as long as the seed, the benchmark and the server see the same ones.
|
||||
|
||||
The seed writes ten years for one user: Apple Health dense rows (heart rate about every ten minutes, HRV, SpO2, respiratory rate, gait, audio), Withings blood pressure, pulse and scale, a weekly manual weight, three years of Oura resting heart rate and RMSSD, 90 days of Nightscout CGM, `stats:` and per-sample cumulative rows, nightly sleep stages, about 3,000 workouts, three years of mood and of three medications with their intake events. It then folds the rollup tier the way the boot backfill does and runs `ANALYZE`. It takes about 80 seconds on the Mac mini.
|
||||
|
||||
The seed is deterministic for a given `--anchor`: values and timestamps are functions of the row index and that day. Without `--anchor` it anchors on today, which is what you want when the benchmark should see recent data.
|
||||
|
||||
### 3. A production build
|
||||
|
||||
```sh
|
||||
pnpm build
|
||||
```
|
||||
|
||||
### 4. Run
|
||||
|
||||
```sh
|
||||
pnpm dlx tsx scripts/bench/run.ts --out scripts/bench/results/<label>
|
||||
```
|
||||
|
||||
The script boots the build once with every role, so the worker creates its queues as it does in production, drops whatever that boot scheduled, and then runs the measurement against a web-only process (`HEALTHLOG_PROCESS_TYPE=web`) under the production heap ceiling (`--max-old-space-size=792`, the value Node derives in the 1.5 GB production container). Without the worker, its polling does not land in every statement count. It prints the server's idle statement rate so the remaining background is visible.
|
||||
|
||||
For each route it sends three warm-up requests, then twenty measured ones in each of two modes, one request at a time:
|
||||
|
||||
- **cold**: before every request a one-row ingest marked as a manual sync evicts the user's caches, so the request rebuilds what it serves. This is a reader's first look after a sync.
|
||||
- **warm**: back to back, what a cache hit costs.
|
||||
|
||||
The batch ingest posts 5,000 fresh heart-rate rows as ten requests of 500, the route's per-request limit, and removes them again at the end.
|
||||
|
||||
Statements are the change in `sum(calls)` of `pg_stat_statements` over the database, read before and after each request, once no statement is in flight and the counter holds still. Fire-and-forget writes a request starts are therefore counted with it.
|
||||
|
||||
Options: `--iterations`, `--warmup`, `--user`, `--heap-mb`, `--port`, and `--base-url` to measure a server you started yourself.
|
||||
|
||||
The output is `<label>.json` (every number, plus the commit, Node and Postgres versions and the row count) and `<label>.md` (the table). Commit the pair when a change is meant to move a number, so the PR shows before and after.
|
||||
|
||||
### Reading the numbers
|
||||
|
||||
Run on a quiet machine; a parallel build or test run moves latency by tens of percent and leaves the statement counts alone. Compare statement counts across machines, latency only across runs on the same machine. A local number is only valid under the production heap ceiling, which the script sets.
|
||||
|
||||
## Statement budgets in the integration suite
|
||||
|
||||
`tests/integration/query-budget-hot-routes.test.ts` seeds a small account (four months of the main types with overlapping sources, sleep, glucose, a medication, workouts and mood), and counts the statements of one request per hot route on a cache miss: `/api/auth/me`, the dashboard summary and snapshot, the pulse series over 30, 365 and 3650 days, the day view, insights health status, correlations, and a 500-row batch ingest.
|
||||
|
||||
The counter (`tests/integration/query-counter.ts`) wraps `pg`'s `Client.prototype.query` on the copy of `pg` the Prisma adapter loads, and settles background tasks before it closes the window.
|
||||
|
||||
Each budget is the measured count. When a change lowers one, lower the budget in the same change. When a change has to raise one, raise it with the reason in the PR. A failure prints every statement the request sent, grouped by text, so an added loop shows up as one statement with a count next to it.
|
||||
|
||||
Run it like any integration file:
|
||||
|
||||
```sh
|
||||
pnpm test:integration tests/integration/query-budget-hot-routes.test.ts
|
||||
```
|
||||
@@ -28,6 +28,7 @@ typecheck.
|
||||
- **`repair-intake-anomalies.ts`** — repair historic medication-ledger defects (duplicate dose-slot rows, implausible `taken_at`); dry-run by default, see `docs/ops/intake-repair.md`.
|
||||
- **`backfill-rollups.ts`**, **`drain-per-sample-cumulative.ts`**, **`backfill-mood-note-column.ts`** — historical data backfills (note: the boot-time `rollup-full-backfill` queue handles new accounts automatically; the CLI is the manual fallback).
|
||||
- **`seed-demo.ts`** + **`seed-demo-v15.sql`** — populate a demo tenant.
|
||||
- **`bench/`** — `seed-large-account.ts` (a synthetic ten-year account of about 1.26 million measurement rows) and `run.ts` (latency and SQL statements per request against it); see `docs/contributing/benchmarks.md`. Results land in `bench/results/`.
|
||||
- **`test-notifications.ts`** — exercise the notification dispatcher + reminder check.
|
||||
- **`fetch-geolite2.sh`** — download the MaxMind GeoLite2 MMDBs (needs a licence key) before `docker build`.
|
||||
- **`install-hooks.sh`** — install the local git hooks.
|
||||
|
||||
@@ -0,0 +1,478 @@
|
||||
/**
|
||||
* Request benchmark against a seeded account: latency (p50, p95, max) and
|
||||
* SQL statements per request for the hot read and write paths.
|
||||
*
|
||||
* DATABASE_URL=postgresql://… API_TOKEN_HMAC_KEY=… ENCRYPTION_KEYS=… \
|
||||
* ENCRYPTION_ACTIVE_KEY_ID=… pnpm dlx tsx scripts/bench/run.ts \
|
||||
* [--user bench-large] [--iterations 20] [--warmup 3] \
|
||||
* [--base-url http://127.0.0.1:3100] [--out scripts/bench/results/<label>]
|
||||
*
|
||||
* See docs/contributing/benchmarks.md for the whole setup (a throwaway
|
||||
* Postgres with pg_stat_statements, migrations, the seed, a production
|
||||
* build). Never point it at a database that holds real data: it mints a
|
||||
* token for the bench user and writes measurements under it.
|
||||
*
|
||||
* Without `--base-url` the script starts the production build itself
|
||||
* (`next start` on `.next`) as a web-only process under the production heap
|
||||
* ceiling, and stops it at the end.
|
||||
*
|
||||
* Statements are counted with pg_stat_statements: the sum of `calls` over
|
||||
* the database, read before and after each request, once no statement is
|
||||
* in flight and the count has stopped moving (so fire-and-forget writes a
|
||||
* request starts are counted with it). The bench's own reads of the two
|
||||
* views are excluded. Requests run one at a time, so nothing else is in the
|
||||
* delta except the server's idle background, which is measured first and
|
||||
* reported. Before the measurement the build boots once with every role, so
|
||||
* the worker creates its queues as it does in production; whatever that
|
||||
* boot scheduled is dropped.
|
||||
*
|
||||
* Each read runs twice: cold, with the user's caches evicted before every
|
||||
* request by a one-row manual-sync ingest (excluded from the numbers), so
|
||||
* the request rebuilds what it serves; and warm, back to back, which is
|
||||
* what a cache hit costs. The batch ingest posts 5,000 fresh rows as ten
|
||||
* requests of 500, the route's per-request limit, and removes them again.
|
||||
*
|
||||
* Output: `<out>.json` and `<out>.md`. Exit 2 on any non-2xx response.
|
||||
*/
|
||||
import { execSync, spawn, type ChildProcess } from "node:child_process";
|
||||
import { mkdirSync, writeFileSync } from "node:fs";
|
||||
import { cpus, loadavg } from "node:os";
|
||||
import { dirname, join } from "node:path";
|
||||
|
||||
import { Pool } from "pg";
|
||||
|
||||
import { issueApiToken } from "@/lib/auth/issue-token";
|
||||
import { prisma } from "@/lib/db";
|
||||
|
||||
function arg(name: string, fallback: string): string {
|
||||
const i = process.argv.indexOf(`--${name}`);
|
||||
return i >= 0 && process.argv[i + 1] ? process.argv[i + 1]! : fallback;
|
||||
}
|
||||
|
||||
const USER_ID = arg("user", "bench-large");
|
||||
const ITERATIONS = Number(arg("iterations", "20"));
|
||||
const WARMUP = Number(arg("warmup", "3"));
|
||||
const PORT = Number(arg("port", "3100"));
|
||||
const BASE_URL = arg("base-url", "");
|
||||
const OUT = arg("out", join("scripts", "bench", "results", "latest"));
|
||||
const HEAP_MB = Number(arg("heap-mb", "792"));
|
||||
|
||||
const pool = new Pool({ connectionString: process.env.DATABASE_URL, max: 1 });
|
||||
|
||||
async function statementCount(): Promise<number> {
|
||||
const { rows } = await pool.query<{ n: string | null }>(
|
||||
`SELECT sum(calls)::bigint AS n FROM pg_stat_statements
|
||||
WHERE dbid = (SELECT oid FROM pg_database WHERE datname = current_database())
|
||||
AND query NOT ILIKE '%pg_stat_statements%'
|
||||
AND query NOT ILIKE '%pg_stat_activity%'`,
|
||||
);
|
||||
return Number(rows[0]?.n ?? 0);
|
||||
}
|
||||
|
||||
const sleep = (ms: number) => new Promise((r) => setTimeout(r, ms));
|
||||
|
||||
/**
|
||||
* Wait until the server has no statement in flight and the counter holds
|
||||
* still for 150 ms (or 10 s pass), then read it. pg_stat_statements counts a
|
||||
* statement when it ends, so a long fold in a background refresh would look
|
||||
* settled by the counter alone; pg_stat_activity says it is still running.
|
||||
*/
|
||||
async function settledCount(): Promise<number> {
|
||||
let last = -1;
|
||||
const deadline = Date.now() + 10_000;
|
||||
while (Date.now() < deadline) {
|
||||
const { rows } = await pool.query<{ active: string }>(
|
||||
`SELECT count(*)::int AS active FROM pg_stat_activity
|
||||
WHERE datname = current_database() AND pid <> pg_backend_pid()
|
||||
AND state NOT IN ('idle')`,
|
||||
);
|
||||
const now = await statementCount();
|
||||
if (Number(rows[0]?.active ?? 0) === 0 && now === last) return now;
|
||||
last = now;
|
||||
await sleep(150);
|
||||
}
|
||||
return last;
|
||||
}
|
||||
|
||||
function percentile(sorted: number[], p: number): number {
|
||||
if (sorted.length === 0) return 0;
|
||||
const i = Math.ceil((p / 100) * sorted.length) - 1;
|
||||
return sorted[Math.min(sorted.length - 1, Math.max(0, i))]!;
|
||||
}
|
||||
|
||||
interface Sample {
|
||||
ms: number;
|
||||
statements: number;
|
||||
status: number;
|
||||
}
|
||||
|
||||
interface ScenarioResult {
|
||||
name: string;
|
||||
mode: "cold" | "warm" | "write";
|
||||
request: string;
|
||||
n: number;
|
||||
p50Ms: number;
|
||||
p95Ms: number;
|
||||
maxMs: number;
|
||||
statementsMedian: number;
|
||||
statementsMax: number;
|
||||
failures: number;
|
||||
}
|
||||
|
||||
let token = "";
|
||||
const failures: string[] = [];
|
||||
|
||||
async function call(
|
||||
method: string,
|
||||
path: string,
|
||||
body?: unknown,
|
||||
): Promise<Sample> {
|
||||
const before = await settledCount();
|
||||
const started = performance.now();
|
||||
const res = await fetch(`${base()}${path}`, {
|
||||
method,
|
||||
headers: {
|
||||
authorization: `Bearer ${token}`,
|
||||
...(body ? { "content-type": "application/json" } : {}),
|
||||
},
|
||||
body: body ? JSON.stringify(body) : undefined,
|
||||
});
|
||||
await res.arrayBuffer();
|
||||
const ms = performance.now() - started;
|
||||
const after = await settledCount();
|
||||
if (!res.ok) failures.push(`${method} ${path}: ${res.status}`);
|
||||
return { ms, statements: after - before, status: res.status };
|
||||
}
|
||||
|
||||
function summarise(
|
||||
name: string,
|
||||
mode: ScenarioResult["mode"],
|
||||
request: string,
|
||||
samples: Sample[],
|
||||
): ScenarioResult {
|
||||
const ms = samples.map((s) => s.ms).sort((a, b) => a - b);
|
||||
const st = samples.map((s) => s.statements).sort((a, b) => a - b);
|
||||
const r = (n: number) => Math.round(n * 10) / 10;
|
||||
return {
|
||||
name,
|
||||
mode,
|
||||
request,
|
||||
n: samples.length,
|
||||
p50Ms: r(percentile(ms, 50)),
|
||||
p95Ms: r(percentile(ms, 95)),
|
||||
maxMs: r(ms[ms.length - 1] ?? 0),
|
||||
statementsMedian: percentile(st, 50),
|
||||
statementsMax: st[st.length - 1] ?? 0,
|
||||
failures: samples.filter((s) => s.status >= 300).length,
|
||||
};
|
||||
}
|
||||
|
||||
/**
|
||||
* The per-user rate limits would refuse a benchmark long before a person
|
||||
* meets them (one run sends several hundred requests in minutes), so the
|
||||
* bench user's buckets are cleared before each scenario. The limiter's own
|
||||
* cost stays in every count.
|
||||
*/
|
||||
async function resetRateLimits(): Promise<void> {
|
||||
await pool.query(`DELETE FROM rate_limits WHERE key LIKE '%:' || $1`, [
|
||||
USER_ID,
|
||||
]);
|
||||
}
|
||||
|
||||
let bustCounter = 0;
|
||||
/**
|
||||
* Evict the user's caches with a real change, outside the numbers: a one-row
|
||||
* ingest marked as a manual sync, which hard-evicts instead of marking
|
||||
* stale, so the next read is a miss and rebuilds in the request.
|
||||
*/
|
||||
async function bust(anchorDay: string): Promise<void> {
|
||||
bustCounter += 1;
|
||||
const res = await fetch(`${base()}/api/measurements/batch`, {
|
||||
method: "POST",
|
||||
headers: {
|
||||
authorization: `Bearer ${token}`,
|
||||
"content-type": "application/json",
|
||||
},
|
||||
body: JSON.stringify({
|
||||
syncTrigger: "manual",
|
||||
entries: [
|
||||
{
|
||||
hkIdentifier: "HKQuantityTypeIdentifierFlightsClimbed",
|
||||
value: 10 + (bustCounter % 2),
|
||||
unit: "count",
|
||||
startDate: `${anchorDay}T08:00:00.000Z`,
|
||||
endDate: `${anchorDay}T18:00:00.000Z`,
|
||||
externalId: `stats:HKQuantityTypeIdentifierFlightsClimbed:${anchorDay}`,
|
||||
},
|
||||
],
|
||||
}),
|
||||
});
|
||||
const body = (await res.json()) as {
|
||||
data?: { inserted?: number; updated?: number };
|
||||
};
|
||||
// A bust that wrote nothing evicted nothing, and every "cold" number
|
||||
// after it would be a cache hit.
|
||||
if ((body.data?.inserted ?? 0) + (body.data?.updated ?? 0) !== 1) {
|
||||
throw new Error(`cache bust wrote no row: ${JSON.stringify(body)}`);
|
||||
}
|
||||
}
|
||||
|
||||
function base(): string {
|
||||
return BASE_URL || `http://127.0.0.1:${PORT}`;
|
||||
}
|
||||
|
||||
async function waitForHealth(
|
||||
timeoutMs: number,
|
||||
server: ChildProcess | null,
|
||||
): Promise<void> {
|
||||
const deadline = Date.now() + timeoutMs;
|
||||
while (Date.now() < deadline) {
|
||||
if (server && server.exitCode !== null) {
|
||||
throw new Error(`the server exited with code ${server.exitCode}`);
|
||||
}
|
||||
try {
|
||||
const res = await fetch(`${base()}/api/health`);
|
||||
if (res.ok) return;
|
||||
} catch {
|
||||
// not listening yet
|
||||
}
|
||||
await sleep(500);
|
||||
}
|
||||
throw new Error(`server at ${base()} did not answer /api/health`);
|
||||
}
|
||||
|
||||
function startServer(role: "web" | "all"): ChildProcess {
|
||||
// `next start` rather than the standalone server: the standalone tree
|
||||
// lacks the externalised runtime packages the image installs separately
|
||||
// (see the Dockerfile), while the checkout has them. It serves the same
|
||||
// build output.
|
||||
const child = spawn(
|
||||
process.execPath,
|
||||
[
|
||||
"node_modules/next/dist/bin/next",
|
||||
"start",
|
||||
"-p",
|
||||
String(PORT),
|
||||
"-H",
|
||||
"127.0.0.1",
|
||||
],
|
||||
{
|
||||
env: {
|
||||
...process.env,
|
||||
NODE_ENV: "production",
|
||||
PORT: String(PORT),
|
||||
HOSTNAME: "127.0.0.1",
|
||||
HEALTHLOG_PROCESS_TYPE: role,
|
||||
SESSION_COOKIE_SECURE: "false",
|
||||
NODE_OPTIONS: `--max-old-space-size=${HEAP_MB}`,
|
||||
},
|
||||
stdio: ["ignore", "ignore", "inherit"],
|
||||
},
|
||||
);
|
||||
return child;
|
||||
}
|
||||
|
||||
async function stopServer(child: ChildProcess): Promise<void> {
|
||||
if (child.exitCode !== null) return;
|
||||
const exited = new Promise((resolve) => child.once("exit", resolve));
|
||||
child.kill("SIGTERM");
|
||||
await Promise.race([exited, sleep(20_000)]);
|
||||
if (child.exitCode === null) child.kill("SIGKILL");
|
||||
}
|
||||
|
||||
function freshBatch(run: string, chunk: number): unknown[] {
|
||||
// 500 heart-rate readings a minute apart, on a day inside the seeded
|
||||
// window, with ids no earlier run used.
|
||||
const day = Date.UTC(2026, 0, 1) + chunk * 86_400_000;
|
||||
return Array.from({ length: 500 }, (_, i) => {
|
||||
const at = new Date(day + i * 60_000).toISOString();
|
||||
return {
|
||||
hkIdentifier: "HKQuantityTypeIdentifierHeartRate",
|
||||
value: 55 + (i % 40),
|
||||
unit: "count/min",
|
||||
startDate: at,
|
||||
endDate: at,
|
||||
externalId: `bench-${run}-${chunk}-${i}`,
|
||||
};
|
||||
});
|
||||
}
|
||||
|
||||
function markdown(meta: Record<string, unknown>, rows: ScenarioResult[]) {
|
||||
const lines = [
|
||||
`# Benchmark ${String(meta.label)}`,
|
||||
"",
|
||||
`Commit ${String(meta.commit)}, ${String(meta.date)}. ${String(meta.measurements)} measurement rows on the bench account, ${ITERATIONS} measured requests per row after ${WARMUP} warm-up requests, one at a time. Server: production build, web process, heap ceiling ${HEAP_MB} MB. Postgres ${String(meta.postgres)}, ${String(meta.cpus)} cores, load average ${String(meta.loadAverage)} at the end. Idle background: ${String(meta.idleStatementsPerSecond)} statements/s.`,
|
||||
"",
|
||||
"| Request | Mode | n | p50 ms | p95 ms | max ms | statements (median) | statements (max) |",
|
||||
"| --- | --- | ---: | ---: | ---: | ---: | ---: | ---: |",
|
||||
...rows.map(
|
||||
(r) =>
|
||||
`| ${r.name} | ${r.mode} | ${r.n} | ${r.p50Ms} | ${r.p95Ms} | ${r.maxMs} | ${r.statementsMedian} | ${r.statementsMax} |`,
|
||||
),
|
||||
"",
|
||||
];
|
||||
return lines.join("\n");
|
||||
}
|
||||
|
||||
async function main(): Promise<void> {
|
||||
if (!process.env.DATABASE_URL) throw new Error("DATABASE_URL is required");
|
||||
await pool.query("CREATE EXTENSION IF NOT EXISTS pg_stat_statements");
|
||||
await statementCount(); // fails here when the library is not preloaded
|
||||
|
||||
const user = await prisma.user.findUnique({
|
||||
where: { id: USER_ID },
|
||||
select: { id: true },
|
||||
});
|
||||
if (!user) throw new Error(`no user ${USER_ID}; run the seed first`);
|
||||
const measurements = await prisma.measurement.count({
|
||||
where: { userId: USER_ID },
|
||||
});
|
||||
token = (
|
||||
await issueApiToken({
|
||||
userId: USER_ID,
|
||||
name: "benchmark",
|
||||
permissions: ["*"],
|
||||
expiresInDays: 1,
|
||||
})
|
||||
).token;
|
||||
|
||||
// A web-only process only sends jobs; the queues are created by the
|
||||
// worker role. Boot the full process once so every queue exists (as it
|
||||
// does in production), then run the measurement against a web-only
|
||||
// process, so the worker's polling is not in every delta.
|
||||
if (!BASE_URL) {
|
||||
const full = startServer("all");
|
||||
await waitForHealth(180_000, full);
|
||||
await stopServer(full);
|
||||
// Drop whatever that boot scheduled: nothing in the web-only process
|
||||
// runs jobs, and a fold left queued would change the fixture on the
|
||||
// next full boot.
|
||||
await pool.query("TRUNCATE pgboss.job");
|
||||
}
|
||||
const server = BASE_URL ? null : startServer("web");
|
||||
try {
|
||||
await waitForHealth(180_000, server);
|
||||
|
||||
const idleBefore = await settledCount();
|
||||
await sleep(5_000);
|
||||
const idleStatementsPerSecond =
|
||||
Math.round(((await statementCount()) - idleBefore) / 0.5) / 10;
|
||||
|
||||
const yesterday = new Date(Date.now() - 86_400_000)
|
||||
.toISOString()
|
||||
.slice(0, 10);
|
||||
const reads: Array<[string, string]> = [
|
||||
["/api/auth/me", "/api/auth/me"],
|
||||
["dashboard summary", "/api/dashboard/summary"],
|
||||
["dashboard snapshot", "/api/dashboard/snapshot"],
|
||||
["series pulse 30 d", "/api/measurements/series?kind=pulse&days=30"],
|
||||
["series pulse 365 d", "/api/measurements/series?kind=pulse&days=365"],
|
||||
["series pulse 3650 d", "/api/measurements/series?kind=pulse&days=3650"],
|
||||
["day view", `/api/day/${yesterday}`],
|
||||
["insights health status", "/api/insights/health-status"],
|
||||
["insights correlations", "/api/insights/correlations"],
|
||||
];
|
||||
|
||||
const results: ScenarioResult[] = [];
|
||||
for (const [name, path] of reads) {
|
||||
await resetRateLimits();
|
||||
for (let i = 0; i < WARMUP; i++) await call("GET", path);
|
||||
const cold: Sample[] = [];
|
||||
for (let i = 0; i < ITERATIONS; i++) {
|
||||
await bust(yesterday);
|
||||
cold.push(await call("GET", path));
|
||||
}
|
||||
results.push(summarise(name, "cold", `GET ${path}`, cold));
|
||||
await resetRateLimits();
|
||||
const warm: Sample[] = [];
|
||||
for (let i = 0; i < ITERATIONS; i++) warm.push(await call("GET", path));
|
||||
results.push(summarise(name, "warm", `GET ${path}`, warm));
|
||||
console.log(
|
||||
`${name.padEnd(26)} cold p95 ${results.at(-2)!.p95Ms} ms / ${results.at(-2)!.statementsMedian} st, warm p95 ${results.at(-1)!.p95Ms} ms / ${results.at(-1)!.statementsMedian} st`,
|
||||
);
|
||||
}
|
||||
|
||||
// Rows an earlier run wrote would turn this run's inserts into
|
||||
// reconciles of the same natural key, so they go first, and again at
|
||||
// the end, leaving the seeded fixture as it was.
|
||||
const clean = () =>
|
||||
pool.query(
|
||||
`DELETE FROM measurements WHERE user_id = $1 AND external_id LIKE 'bench-%'`,
|
||||
[USER_ID],
|
||||
);
|
||||
await clean();
|
||||
await resetRateLimits();
|
||||
const run = Date.now().toString(36);
|
||||
const writes: Sample[] = [];
|
||||
for (let chunk = 0; chunk < 10; chunk++) {
|
||||
writes.push(
|
||||
await call("POST", "/api/measurements/batch", {
|
||||
entries: freshBatch(run, chunk),
|
||||
}),
|
||||
);
|
||||
}
|
||||
const batch = summarise(
|
||||
"batch ingest 5k (per 500)",
|
||||
"write",
|
||||
"POST /api/measurements/batch ×10",
|
||||
writes,
|
||||
);
|
||||
results.push(batch);
|
||||
await clean();
|
||||
const totalMs = writes.reduce((a, s) => a + s.ms, 0);
|
||||
const totalStatements = writes.reduce((a, s) => a + s.statements, 0);
|
||||
console.log(
|
||||
`batch ingest 5k total ${Math.round(totalMs)} ms / ${totalStatements} st`,
|
||||
);
|
||||
|
||||
let commit = "unknown";
|
||||
try {
|
||||
commit = execSync("git rev-parse --short HEAD").toString().trim();
|
||||
} catch {
|
||||
// not a checkout
|
||||
}
|
||||
const meta = {
|
||||
label: OUT.split("/").pop(),
|
||||
commit,
|
||||
date: new Date().toISOString(),
|
||||
node: process.version,
|
||||
postgres: (
|
||||
await pool.query<{ v: string }>(
|
||||
"SELECT current_setting('server_version') AS v",
|
||||
)
|
||||
).rows[0]?.v,
|
||||
cpus: cpus().length,
|
||||
loadAverage: loadavg().map((n) => Math.round(n * 10) / 10),
|
||||
measurements,
|
||||
iterations: ITERATIONS,
|
||||
warmup: WARMUP,
|
||||
heapMb: HEAP_MB,
|
||||
idleStatementsPerSecond,
|
||||
batch5k: { totalMs: Math.round(totalMs), totalStatements },
|
||||
};
|
||||
mkdirSync(dirname(OUT), { recursive: true });
|
||||
writeFileSync(
|
||||
`${OUT}.json`,
|
||||
`${JSON.stringify({ meta, results }, null, 2)}\n`,
|
||||
);
|
||||
writeFileSync(`${OUT}.md`, markdown(meta, results));
|
||||
console.log(`wrote ${OUT}.json and ${OUT}.md`);
|
||||
} finally {
|
||||
if (server) await stopServer(server);
|
||||
}
|
||||
|
||||
if (failures.length > 0) {
|
||||
console.error(`non-2xx responses:\n ${failures.join("\n ")}`);
|
||||
process.exitCode = 2;
|
||||
}
|
||||
}
|
||||
|
||||
main()
|
||||
.catch((err) => {
|
||||
console.error(err);
|
||||
process.exitCode = 1;
|
||||
})
|
||||
.finally(async () => {
|
||||
await prisma.$disconnect();
|
||||
await pool.end();
|
||||
});
|
||||
@@ -0,0 +1,659 @@
|
||||
/**
|
||||
* Seed one synthetic account the size of a heavy real one, for the benchmark
|
||||
* harness (`scripts/bench/run.ts`) and for reproducing the maintenance jobs:
|
||||
* about 1.26 million measurement rows over ten years, plus workouts, mood
|
||||
* entries and medications with their intake history.
|
||||
*
|
||||
* DATABASE_URL=postgresql://… pnpm dlx tsx scripts/bench/seed-large-account.ts \
|
||||
* [--user bench-large] [--years 10] [--anchor 2026-10-22] [--reset]
|
||||
* [--no-rollups]
|
||||
*
|
||||
* Never point this at a database that holds real data: it inserts rows under
|
||||
* the given user id and, with `--reset`, deletes that user (and, through the
|
||||
* cascades, everything it owns) first.
|
||||
*
|
||||
* Deterministic: every value and timestamp is a function of the row index and
|
||||
* the anchor day (`--anchor`, UTC midnight; default today), and ids are
|
||||
* derived from the user id, the series and the index. Two runs with the same
|
||||
* arguments write the same rows.
|
||||
*
|
||||
* The shape follows what the integrations write, with overlapping sources so
|
||||
* the canonical-source pick has work to do:
|
||||
*
|
||||
* - Apple Health dense spot rows (heart rate about every ten minutes, HRV,
|
||||
* SpO2, respiratory rate, gait metrics, audio exposure) keyed by a hashed
|
||||
* sample id;
|
||||
* - Withings morning and evening blood pressure with its pulse, and daily
|
||||
* weight, body fat and muscle mass; a weekly manual weight beside it;
|
||||
* - Oura nightly resting heart rate and RMSSD for the last three years,
|
||||
* beside Apple Health's resting heart rate;
|
||||
* - Nightscout CGM glucose every five minutes for the last 90 days;
|
||||
* - one `stats:` daily row per cumulative type and day (the export import)
|
||||
* and per-sample cumulative rows with opaque ids for the last 90 days
|
||||
* (the iOS sync);
|
||||
* - sleep-stage segments every night;
|
||||
* - about 3,000 workouts, a mood entry a day for three years, and three
|
||||
* daily medications with their intake events for three years.
|
||||
*
|
||||
* Everything goes in through set-based `INSERT … SELECT … generate_series`,
|
||||
* so a million rows take seconds rather than a million round trips. The
|
||||
* rollup tier is then folded with `recomputeUserRollups`, the same path the
|
||||
* boot backfill takes for an imported account, so reads see the state a real
|
||||
* account is in. `ANALYZE` runs at the end: a planner reading pre-seed
|
||||
* statistics picks plans no production table of this size would get.
|
||||
*/
|
||||
import { prisma } from "@/lib/db";
|
||||
import {
|
||||
ALL_GRANULARITIES,
|
||||
recomputeUserRollups,
|
||||
} from "@/lib/rollups/measurement-rollups";
|
||||
|
||||
function arg(name: string, fallback: string): string {
|
||||
const i = process.argv.indexOf(`--${name}`);
|
||||
return i >= 0 && process.argv[i + 1] ? process.argv[i + 1]! : fallback;
|
||||
}
|
||||
|
||||
const USER_ID = arg("user", "bench-large");
|
||||
const YEARS = Number(arg("years", "10"));
|
||||
const ANCHOR = arg("anchor", new Date().toISOString().slice(0, 10));
|
||||
const RESET = process.argv.includes("--reset");
|
||||
const ROLLUPS = !process.argv.includes("--no-rollups");
|
||||
const ZONE = "Europe/Berlin";
|
||||
|
||||
if (!/^\d{4}-\d{2}-\d{2}$/.test(ANCHOR)) {
|
||||
throw new Error(`--anchor must be YYYY-MM-DD, got ${ANCHOR}`);
|
||||
}
|
||||
|
||||
/**
|
||||
* A spot series: rows evenly spaced over the last `days` days before the
|
||||
* anchor. `perDay` above one spaces rows through the day; `everyDays` above
|
||||
* one writes one row every that many days at `offsetSec` past midnight UTC.
|
||||
*/
|
||||
interface SpotSeries {
|
||||
type: string;
|
||||
unit: string;
|
||||
source: string;
|
||||
base: number;
|
||||
spread: number;
|
||||
days: number;
|
||||
perDay?: number;
|
||||
everyDays?: number;
|
||||
offsetSec?: number;
|
||||
/** externalId prefix; `stats:` rows are generated separately. */
|
||||
idPrefix: string;
|
||||
}
|
||||
|
||||
async function insertSpot(s: SpotSeries): Promise<number> {
|
||||
const step = s.everyDays
|
||||
? s.everyDays * 86_400
|
||||
: Math.floor(86_400 / (s.perDay ?? 1));
|
||||
const count = Math.floor((s.days * 86_400) / step);
|
||||
// Jitter stays below the step so every measured_at is distinct per type
|
||||
// and source (the (user, type, measured_at, source, sleep_stage) unique
|
||||
// index). Series that share a step and offset (BP systolic, diastolic and
|
||||
// the Withings pulse) share their timestamps, as a cuff reading does.
|
||||
const jitter = Math.max(1, Math.min(59, step - 1));
|
||||
return prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-' || $2 || '-' || $10 || '-' || g,
|
||||
$1,
|
||||
$2::measurement_type,
|
||||
round(($3 + $4 * sin(g / 37.0) + ($4 / 2) * ((g * 7919) % 100) / 100.0)::numeric, 2),
|
||||
$5,
|
||||
$10::measurement_source,
|
||||
$11::date::timestamp - make_interval(secs => $6 * $7::bigint)
|
||||
+ make_interval(secs => g * $6 + $12 + (g * 31) % $8),
|
||||
$9 || md5($2 || $10 || g::text),
|
||||
now(), now()
|
||||
FROM generate_series(0, $7::bigint - 1) AS g
|
||||
`,
|
||||
USER_ID,
|
||||
s.type,
|
||||
s.base,
|
||||
s.spread,
|
||||
s.unit,
|
||||
step,
|
||||
count,
|
||||
jitter,
|
||||
s.idPrefix,
|
||||
s.source,
|
||||
ANCHOR,
|
||||
s.offsetSec ?? 0,
|
||||
);
|
||||
}
|
||||
|
||||
function log(label: string, n: number): void {
|
||||
console.log(`${label.padEnd(40)} ${String(n).padStart(9)}`);
|
||||
}
|
||||
|
||||
async function seedMeasurements(days: number): Promise<number> {
|
||||
const recent = Math.min(days, 90);
|
||||
const threeYears = Math.min(days, 3 * 365);
|
||||
const spot: SpotSeries[] = [
|
||||
// Apple Health, the dense tier.
|
||||
{
|
||||
type: "PULSE",
|
||||
unit: "bpm",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 147,
|
||||
base: 62,
|
||||
spread: 25,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "HEART_RATE_VARIABILITY",
|
||||
unit: "ms",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 12,
|
||||
base: 45,
|
||||
spread: 20,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "OXYGEN_SATURATION",
|
||||
unit: "%",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 24,
|
||||
base: 96,
|
||||
spread: 2,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "RESPIRATORY_RATE",
|
||||
unit: "breaths/min",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 24,
|
||||
base: 14,
|
||||
spread: 3,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_SPEED",
|
||||
unit: "m/s",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 12,
|
||||
base: 1.3,
|
||||
spread: 0.2,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_STEP_LENGTH",
|
||||
unit: "m",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 12,
|
||||
base: 0.72,
|
||||
spread: 0.06,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_ASYMMETRY",
|
||||
unit: "%",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 12,
|
||||
base: 3,
|
||||
spread: 2,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_DOUBLE_SUPPORT",
|
||||
unit: "%",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 12,
|
||||
base: 28,
|
||||
spread: 3,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "AUDIO_EXPOSURE_ENV",
|
||||
unit: "dBA",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 24,
|
||||
base: 62,
|
||||
spread: 12,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "RESTING_HEART_RATE",
|
||||
unit: "bpm",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 1,
|
||||
offsetSec: 6 * 3600,
|
||||
base: 56,
|
||||
spread: 4,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "TIME_IN_DAYLIGHT",
|
||||
unit: "minutes",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 1,
|
||||
offsetSec: 18 * 3600,
|
||||
base: 60,
|
||||
spread: 40,
|
||||
days,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
// The iOS sync's per-sample cumulative rows: dense, recent, opaque ids.
|
||||
{
|
||||
type: "ACTIVITY_STEPS",
|
||||
unit: "steps",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 100,
|
||||
base: 90,
|
||||
spread: 60,
|
||||
days: recent,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
{
|
||||
type: "ACTIVE_ENERGY_BURNED",
|
||||
unit: "kcal",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 400,
|
||||
base: 1.2,
|
||||
spread: 0.9,
|
||||
days: recent,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_RUNNING_DISTANCE",
|
||||
unit: "m",
|
||||
source: "APPLE_HEALTH",
|
||||
perDay: 100,
|
||||
base: 70,
|
||||
spread: 50,
|
||||
days: recent,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
// Withings: a cuff reading morning and evening, and the scale.
|
||||
{
|
||||
type: "BLOOD_PRESSURE_SYS",
|
||||
unit: "mmHg",
|
||||
source: "WITHINGS",
|
||||
perDay: 2,
|
||||
offsetSec: 6 * 3600,
|
||||
base: 126,
|
||||
spread: 10,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
{
|
||||
type: "BLOOD_PRESSURE_DIA",
|
||||
unit: "mmHg",
|
||||
source: "WITHINGS",
|
||||
perDay: 2,
|
||||
offsetSec: 6 * 3600,
|
||||
base: 81,
|
||||
spread: 7,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
{
|
||||
type: "PULSE",
|
||||
unit: "bpm",
|
||||
source: "WITHINGS",
|
||||
perDay: 2,
|
||||
offsetSec: 6 * 3600,
|
||||
base: 66,
|
||||
spread: 8,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
{
|
||||
type: "WEIGHT",
|
||||
unit: "kg",
|
||||
source: "WITHINGS",
|
||||
perDay: 1,
|
||||
offsetSec: 5 * 3600,
|
||||
base: 82,
|
||||
spread: 1.5,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
{
|
||||
type: "BODY_FAT",
|
||||
unit: "%",
|
||||
source: "WITHINGS",
|
||||
perDay: 1,
|
||||
offsetSec: 5 * 3600,
|
||||
base: 22,
|
||||
spread: 2,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
{
|
||||
type: "MUSCLE_MASS",
|
||||
unit: "kg",
|
||||
source: "WITHINGS",
|
||||
perDay: 1,
|
||||
offsetSec: 5 * 3600,
|
||||
base: 61,
|
||||
spread: 1,
|
||||
days,
|
||||
idPrefix: "wi-",
|
||||
},
|
||||
// A weekly manual weight beside the scale.
|
||||
{
|
||||
type: "WEIGHT",
|
||||
unit: "kg",
|
||||
source: "MANUAL",
|
||||
everyDays: 7,
|
||||
offsetSec: 19 * 3600,
|
||||
base: 82.4,
|
||||
spread: 1.5,
|
||||
days,
|
||||
idPrefix: "man-",
|
||||
},
|
||||
// Oura, nightly, the last three years.
|
||||
{
|
||||
type: "RESTING_HEART_RATE",
|
||||
unit: "bpm",
|
||||
source: "OURA",
|
||||
perDay: 1,
|
||||
offsetSec: 4 * 3600,
|
||||
base: 54,
|
||||
spread: 4,
|
||||
days: threeYears,
|
||||
idPrefix: "oura-",
|
||||
},
|
||||
{
|
||||
type: "HRV_RMSSD",
|
||||
unit: "ms",
|
||||
source: "OURA",
|
||||
perDay: 1,
|
||||
offsetSec: 4 * 3600,
|
||||
base: 48,
|
||||
spread: 15,
|
||||
days: threeYears,
|
||||
idPrefix: "oura-",
|
||||
},
|
||||
// Nightscout CGM, every five minutes, the last 90 days.
|
||||
{
|
||||
type: "BLOOD_GLUCOSE",
|
||||
unit: "mg/dL",
|
||||
source: "NIGHTSCOUT",
|
||||
perDay: 288,
|
||||
base: 105,
|
||||
spread: 30,
|
||||
days: recent,
|
||||
idPrefix: "ns-",
|
||||
},
|
||||
];
|
||||
|
||||
let total = 0;
|
||||
for (const s of spot) {
|
||||
const n = await insertSpot(s);
|
||||
total += n;
|
||||
log(`${s.type} (${s.source})`, n);
|
||||
}
|
||||
|
||||
// One export-estimate `stats:` row per cumulative type and local day, at
|
||||
// local noon — what `flushCumulativeBuckets` in the export import writes —
|
||||
// for every day older than the per-sample window.
|
||||
for (const [type, hk, unit, base] of [
|
||||
["ACTIVITY_STEPS", "HKQuantityTypeIdentifierStepCount", "steps", 8000],
|
||||
[
|
||||
"ACTIVE_ENERGY_BURNED",
|
||||
"HKQuantityTypeIdentifierActiveEnergyBurned",
|
||||
"kcal",
|
||||
550,
|
||||
],
|
||||
[
|
||||
"WALKING_RUNNING_DISTANCE",
|
||||
"HKQuantityTypeIdentifierDistanceWalkingRunning",
|
||||
"m",
|
||||
6000,
|
||||
],
|
||||
["FLIGHTS_CLIMBED", "HKQuantityTypeIdentifierFlightsClimbed", "flights", 9],
|
||||
] as const) {
|
||||
const n = await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id,
|
||||
aggregation_provenance, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-' || $2 || '-stats-' || d,
|
||||
$1, $2::measurement_type, $5 + (d * 7919) % ($5 / 2), $4, 'APPLE_HEALTH',
|
||||
(($7::date - d)::timestamp + interval '12 hours')
|
||||
AT TIME ZONE $8 AT TIME ZONE 'UTC',
|
||||
'stats:' || $3 || ':' || to_char($7::date - d, 'YYYY-MM-DD'),
|
||||
'EXPORT_XML_SOURCE_MAX', now(), now()
|
||||
FROM generate_series(91, $6::int) AS d
|
||||
`,
|
||||
USER_ID,
|
||||
type,
|
||||
hk,
|
||||
unit,
|
||||
base,
|
||||
days,
|
||||
ANCHOR,
|
||||
ZONE,
|
||||
);
|
||||
total += n;
|
||||
log(`${type} (stats:)`, n);
|
||||
}
|
||||
|
||||
// Sleep: 30 segments a night from 23:00 local, 16 minutes apart, rotating
|
||||
// through the stages. Distinct measured_at per segment.
|
||||
const sleep = await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id,
|
||||
sleep_stage, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-sleep-' || d || '-' || s,
|
||||
$1, 'SLEEP_DURATION', 16, 'minutes', 'APPLE_HEALTH',
|
||||
(($3::date - d)::timestamp - interval '1 hour' + s * interval '16 minutes')
|
||||
AT TIME ZONE $4 AT TIME ZONE 'UTC',
|
||||
'hk-' || md5('sleep' || d || '-' || s),
|
||||
(ARRAY['CORE','DEEP','CORE','REM','AWAKE','CORE']::sleep_stage[])[1 + s % 6],
|
||||
now(), now()
|
||||
FROM generate_series(1, $2::int) AS d, generate_series(0, 29) AS s
|
||||
`,
|
||||
USER_ID,
|
||||
days,
|
||||
ANCHOR,
|
||||
ZONE,
|
||||
);
|
||||
total += sleep;
|
||||
log("SLEEP_DURATION (APPLE_HEALTH)", sleep);
|
||||
return total;
|
||||
}
|
||||
|
||||
async function seedWorkouts(days: number): Promise<number> {
|
||||
// About one workout every 29 hours, rotating through five sports.
|
||||
const count = Math.floor((days * 24) / 29);
|
||||
return prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO workouts
|
||||
(id, user_id, sport_type, started_at, ended_at, duration_sec,
|
||||
total_energy_kcal, total_distance_m, avg_heart_rate, max_heart_rate,
|
||||
source, external_id, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-workout-' || g,
|
||||
$1,
|
||||
(ARRAY['walking','running','cycling','hiking','swimming'])[1 + g % 5],
|
||||
$3::date::timestamp - make_interval(hours => (g * 29)::int) + interval '17 hours',
|
||||
$3::date::timestamp - make_interval(hours => (g * 29)::int) + interval '17 hours'
|
||||
+ make_interval(secs => (1800 + (g * 97) % 3600)::int),
|
||||
(1800 + (g * 97) % 3600)::int,
|
||||
round((250 + (g * 13) % 400)::numeric, 1),
|
||||
round((3000 + (g * 211) % 9000)::numeric, 1),
|
||||
(115 + (g * 7) % 40)::int,
|
||||
(150 + (g * 5) % 35)::int,
|
||||
'APPLE_HEALTH',
|
||||
'hk-workout-' || md5($1 || g::text),
|
||||
now(), now()
|
||||
FROM generate_series(0, $2::int - 1) AS g
|
||||
`,
|
||||
USER_ID,
|
||||
count,
|
||||
ANCHOR,
|
||||
);
|
||||
}
|
||||
|
||||
async function seedMood(days: number): Promise<number> {
|
||||
const span = Math.min(days, 3 * 365);
|
||||
return prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO mood_entries
|
||||
(id, user_id, date, mood, score, mood_a1, stress_a2, energy_a3,
|
||||
source, mood_logged_at, tz, synced_at, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-mood-' || d,
|
||||
$1,
|
||||
to_char($3::date - d, 'YYYY-MM-DD'),
|
||||
(ARRAY['LAUSIG','SCHLECHT','OKAY','GUT','SUPER_GUT'])[1 + (d * 7) % 5],
|
||||
1 + (d * 7) % 5,
|
||||
2 * ((d * 7) % 5) + 1,
|
||||
(d * 3) % 11,
|
||||
(d * 5) % 11,
|
||||
'MANUAL',
|
||||
(($3::date - d)::timestamp + interval '20 hours') AT TIME ZONE $4 AT TIME ZONE 'UTC',
|
||||
$4,
|
||||
now(), now(), now()
|
||||
FROM generate_series(1, $2::int) AS d
|
||||
`,
|
||||
USER_ID,
|
||||
span,
|
||||
ANCHOR,
|
||||
ZONE,
|
||||
);
|
||||
}
|
||||
|
||||
async function seedMedications(days: number): Promise<number> {
|
||||
const span = Math.min(days, 3 * 365);
|
||||
const meds = [
|
||||
{ key: "a", name: "Synthetic A", dose: "5 mg", window: ["08:00", "09:00"] },
|
||||
{
|
||||
key: "b",
|
||||
name: "Synthetic B",
|
||||
dose: "100 mg",
|
||||
window: ["08:00", "09:00"],
|
||||
},
|
||||
{
|
||||
key: "c",
|
||||
name: "Synthetic C",
|
||||
dose: "20 mg",
|
||||
window: ["20:00", "21:00"],
|
||||
},
|
||||
];
|
||||
let events = 0;
|
||||
for (const m of meds) {
|
||||
const medId = `seed-${USER_ID}-med-${m.key}`;
|
||||
await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO medications
|
||||
(id, user_id, name, dose, starts_on, created_at, updated_at)
|
||||
VALUES ($1, $2, $3, $4, ($5::date - $6::int)::timestamp, now(), now())
|
||||
`,
|
||||
medId,
|
||||
USER_ID,
|
||||
m.name,
|
||||
m.dose,
|
||||
ANCHOR,
|
||||
span,
|
||||
);
|
||||
await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO medication_schedules
|
||||
(id, medication_id, window_start, window_end, schedule_type)
|
||||
VALUES ($1, $2, $3, $4, 'SCHEDULED')
|
||||
`,
|
||||
`${medId}-schedule`,
|
||||
medId,
|
||||
m.window[0],
|
||||
m.window[1],
|
||||
);
|
||||
// One intake event per day; about one in twelve skipped, the rest taken
|
||||
// a few minutes into the window.
|
||||
events += await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO medication_intake_events
|
||||
(id, user_id, medication_id, scheduled_for, taken_at, skipped,
|
||||
attribution_source, source, created_at, updated_at)
|
||||
SELECT
|
||||
$1 || '-intake-' || d,
|
||||
$2,
|
||||
$1,
|
||||
(($3::date - d)::timestamp + $5::time) AT TIME ZONE $6 AT TIME ZONE 'UTC',
|
||||
CASE WHEN d % 12 = 0 THEN NULL
|
||||
ELSE (($3::date - d)::timestamp + $5::time
|
||||
+ make_interval(mins => (d * 7) % 40)) AT TIME ZONE $6 AT TIME ZONE 'UTC'
|
||||
END,
|
||||
d % 12 = 0,
|
||||
'AUTO',
|
||||
'WEB',
|
||||
now(), now()
|
||||
FROM generate_series(1, $4::int) AS d
|
||||
`,
|
||||
medId,
|
||||
USER_ID,
|
||||
ANCHOR,
|
||||
span,
|
||||
m.window[0],
|
||||
ZONE,
|
||||
);
|
||||
}
|
||||
return events;
|
||||
}
|
||||
|
||||
async function main(): Promise<void> {
|
||||
const days = Math.round(YEARS * 365);
|
||||
const started = Date.now();
|
||||
|
||||
if (RESET) {
|
||||
await prisma.$executeRawUnsafe(`DELETE FROM users WHERE id = $1`, USER_ID);
|
||||
}
|
||||
await prisma.$executeRawUnsafe(
|
||||
`INSERT INTO users (id, username, timezone, onboarding_completed_at, updated_at)
|
||||
VALUES ($1, $1, $2, now(), now())
|
||||
ON CONFLICT (id) DO NOTHING`,
|
||||
USER_ID,
|
||||
ZONE,
|
||||
);
|
||||
|
||||
const measurements = await seedMeasurements(days);
|
||||
log("measurements total", measurements);
|
||||
log("workouts", await seedWorkouts(days));
|
||||
log("mood entries", await seedMood(days));
|
||||
log("medication intake events", await seedMedications(days));
|
||||
|
||||
if (ROLLUPS) {
|
||||
const folded = await recomputeUserRollups(USER_ID, {
|
||||
granularities: ALL_GRANULARITIES,
|
||||
statementTimeoutMs: 600_000,
|
||||
});
|
||||
log("rollup rows", folded.rowsUpserted);
|
||||
}
|
||||
|
||||
for (const table of [
|
||||
"measurements",
|
||||
"measurement_rollups",
|
||||
"workouts",
|
||||
"mood_entries",
|
||||
"medication_intake_events",
|
||||
]) {
|
||||
await prisma.$executeRawUnsafe(`ANALYZE ${table}`);
|
||||
}
|
||||
console.log(`done in ${((Date.now() - started) / 1000).toFixed(1)} s`);
|
||||
}
|
||||
|
||||
main()
|
||||
.catch((err) => {
|
||||
console.error(err);
|
||||
process.exitCode = 1;
|
||||
})
|
||||
.finally(() => prisma.$disconnect());
|
||||
@@ -1,296 +0,0 @@
|
||||
/**
|
||||
* Seed one account shaped like a multi-year Apple Health `export.xml` import,
|
||||
* for reproducing the maintenance jobs at the size a real heavy account
|
||||
* reaches (issue #1031: about 1.26 million measurement rows).
|
||||
*
|
||||
* DATABASE_URL=postgresql://… pnpm dlx tsx scripts/seed-large-account.ts \
|
||||
* [--user bench-large] [--years 4] [--pulse-seconds 165]
|
||||
*
|
||||
* Never point this at a database that holds real data: it inserts rows under
|
||||
* the given user id and, with `--reset`, deletes that user first.
|
||||
*
|
||||
* The shape follows what the import and the iOS sync actually write:
|
||||
*
|
||||
* - dense spot rows (heart rate every `--pulse-seconds`, HRV, SpO2) keyed by
|
||||
* a hashed sample id, the tier the dense-intraday retention folds;
|
||||
* - high-frequency mean rows (respiratory rate, gait metrics, audio
|
||||
* exposure), the tier the daily-mean consolidation folds;
|
||||
* - one `stats:` daily row per cumulative type and day (what the export
|
||||
* import writes for steps, energy, distance and flights);
|
||||
* - per-sample cumulative rows with opaque sample ids for the most recent
|
||||
* 90 days (what the iOS sync writes, and what the cumulative drain folds);
|
||||
* - sleep-stage segments every night.
|
||||
*
|
||||
* Everything is generated inside Postgres with `generate_series`, so a million
|
||||
* rows take seconds rather than a million round trips. `ANALYZE` runs at the
|
||||
* end: a planner reading pre-seed statistics picks plans no production table
|
||||
* of this size would get.
|
||||
*/
|
||||
import { prisma } from "@/lib/db";
|
||||
|
||||
function arg(name: string, fallback: string): string {
|
||||
const i = process.argv.indexOf(`--${name}`);
|
||||
return i >= 0 && process.argv[i + 1] ? process.argv[i + 1]! : fallback;
|
||||
}
|
||||
|
||||
const USER_ID = arg("user", "bench-large");
|
||||
const YEARS = Number(arg("years", "4"));
|
||||
const PULSE_SECONDS = Number(arg("pulse-seconds", "165"));
|
||||
const RESET = process.argv.includes("--reset");
|
||||
|
||||
/** A spot series: `perDay` evenly spaced rows per day over the whole span. */
|
||||
interface SpotSeries {
|
||||
type: string;
|
||||
unit: string;
|
||||
perDay: number;
|
||||
base: number;
|
||||
spread: number;
|
||||
/** Days back from now the series covers. Defaults to the whole span. */
|
||||
days?: number;
|
||||
/** externalId prefix; `stats:` rows are generated separately. */
|
||||
idPrefix: string;
|
||||
}
|
||||
|
||||
async function insertSpot(s: SpotSeries, days: number): Promise<number> {
|
||||
const step = Math.floor(86_400 / s.perDay);
|
||||
const count = Math.floor((days * 86_400) / step);
|
||||
// Jitter stays below the step so every measured_at is distinct per type
|
||||
// (the (user, type, measured_at, source, sleep_stage) unique index).
|
||||
const jitter = Math.max(1, Math.min(59, step - 1));
|
||||
return prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-' || $2 || '-' || g,
|
||||
$1,
|
||||
$2::measurement_type,
|
||||
round(($3 + $4 * sin(g / 37.0) + ($4 / 2) * ((g * 7919) % 100) / 100.0)::numeric, 2),
|
||||
$5,
|
||||
'APPLE_HEALTH',
|
||||
date_trunc('minute', now()) - make_interval(secs => $6 * $7::bigint)
|
||||
+ make_interval(secs => g * $6 + (g * 31) % $8),
|
||||
$9 || md5($2 || g::text),
|
||||
now(), now()
|
||||
FROM generate_series(0, $7::bigint - 1) AS g
|
||||
`,
|
||||
USER_ID,
|
||||
s.type,
|
||||
s.base,
|
||||
s.spread,
|
||||
s.unit,
|
||||
step,
|
||||
count,
|
||||
jitter,
|
||||
s.idPrefix,
|
||||
);
|
||||
}
|
||||
|
||||
async function main(): Promise<void> {
|
||||
const days = Math.round(YEARS * 365);
|
||||
|
||||
if (RESET) {
|
||||
await prisma.$executeRawUnsafe(`DELETE FROM users WHERE id = $1`, USER_ID);
|
||||
}
|
||||
await prisma.$executeRawUnsafe(
|
||||
`INSERT INTO users (id, username, timezone, updated_at)
|
||||
VALUES ($1, $1, 'Europe/Berlin', now())
|
||||
ON CONFLICT (id) DO NOTHING`,
|
||||
USER_ID,
|
||||
);
|
||||
|
||||
const pulsePerDay = Math.floor(86_400 / PULSE_SECONDS);
|
||||
const spot: SpotSeries[] = [
|
||||
{
|
||||
type: "PULSE",
|
||||
unit: "bpm",
|
||||
perDay: pulsePerDay,
|
||||
base: 62,
|
||||
spread: 25,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "HEART_RATE_VARIABILITY",
|
||||
unit: "ms",
|
||||
perDay: 12,
|
||||
base: 45,
|
||||
spread: 20,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "OXYGEN_SATURATION",
|
||||
unit: "%",
|
||||
perDay: 24,
|
||||
base: 96,
|
||||
spread: 2,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "RESPIRATORY_RATE",
|
||||
unit: "breaths/min",
|
||||
perDay: 48,
|
||||
base: 14,
|
||||
spread: 3,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_SPEED",
|
||||
unit: "km/h",
|
||||
perDay: 30,
|
||||
base: 4.8,
|
||||
spread: 0.8,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_STEP_LENGTH",
|
||||
unit: "cm",
|
||||
perDay: 30,
|
||||
base: 72,
|
||||
spread: 6,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_ASYMMETRY",
|
||||
unit: "%",
|
||||
perDay: 30,
|
||||
base: 3,
|
||||
spread: 2,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_DOUBLE_SUPPORT",
|
||||
unit: "%",
|
||||
perDay: 30,
|
||||
base: 28,
|
||||
spread: 3,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
{
|
||||
type: "AUDIO_EXPOSURE_ENV",
|
||||
unit: "dB",
|
||||
perDay: 60,
|
||||
base: 62,
|
||||
spread: 12,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
// The iOS sync's per-sample cumulative rows: dense, recent, opaque ids.
|
||||
{
|
||||
type: "ACTIVITY_STEPS",
|
||||
unit: "steps",
|
||||
perDay: 100,
|
||||
base: 90,
|
||||
spread: 60,
|
||||
days: 90,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
{
|
||||
type: "ACTIVE_ENERGY_BURNED",
|
||||
unit: "kcal",
|
||||
perDay: 400,
|
||||
base: 1.2,
|
||||
spread: 0.9,
|
||||
days: 90,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
{
|
||||
type: "WALKING_RUNNING_DISTANCE",
|
||||
unit: "m",
|
||||
perDay: 100,
|
||||
base: 70,
|
||||
spread: 50,
|
||||
days: 90,
|
||||
idPrefix: "uuid-",
|
||||
},
|
||||
{
|
||||
type: "WEIGHT",
|
||||
unit: "kg",
|
||||
perDay: 1,
|
||||
base: 82,
|
||||
spread: 1.5,
|
||||
idPrefix: "hk-",
|
||||
},
|
||||
];
|
||||
|
||||
let total = 0;
|
||||
for (const s of spot) {
|
||||
const n = await insertSpot(s, s.days ?? days);
|
||||
total += n;
|
||||
console.log(`${s.type.padEnd(28)} ${String(n).padStart(9)}`);
|
||||
}
|
||||
|
||||
// One export-estimate `stats:` row per cumulative type and local day, at
|
||||
// local noon — what `flushCumulativeBuckets` in the export import writes.
|
||||
for (const [type, hk, unit, base] of [
|
||||
["ACTIVITY_STEPS", "HKQuantityTypeIdentifierStepCount", "steps", 8000],
|
||||
[
|
||||
"ACTIVE_ENERGY_BURNED",
|
||||
"HKQuantityTypeIdentifierActiveEnergyBurned",
|
||||
"kcal",
|
||||
550,
|
||||
],
|
||||
[
|
||||
"WALKING_RUNNING_DISTANCE",
|
||||
"HKQuantityTypeIdentifierDistanceWalkingRunning",
|
||||
"m",
|
||||
6000,
|
||||
],
|
||||
["FLIGHTS_CLIMBED", "HKQuantityTypeIdentifierFlightsClimbed", "count", 9],
|
||||
] as const) {
|
||||
const n = await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id,
|
||||
aggregation_provenance, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-' || $2 || '-stats-' || d,
|
||||
$1, $2::measurement_type, $5 + (d * 7919) % ($5 / 2), $4, 'APPLE_HEALTH',
|
||||
((current_date - d)::timestamp + interval '12 hours')
|
||||
AT TIME ZONE 'Europe/Berlin' AT TIME ZONE 'UTC',
|
||||
'stats:' || $3 || ':' || to_char(current_date - d, 'YYYY-MM-DD'),
|
||||
'EXPORT_XML_SOURCE_MAX', now(), now()
|
||||
FROM generate_series(91, $6::int) AS d
|
||||
`,
|
||||
USER_ID,
|
||||
type,
|
||||
hk,
|
||||
unit,
|
||||
base,
|
||||
days,
|
||||
);
|
||||
total += n;
|
||||
console.log(`${`${type} (stats:)`.padEnd(28)} ${String(n).padStart(9)}`);
|
||||
}
|
||||
|
||||
// Sleep: 30 segments a night from 23:00 local, 16 minutes apart, rotating
|
||||
// through the stages. Distinct measured_at per segment.
|
||||
const sleep = await prisma.$executeRawUnsafe(
|
||||
`
|
||||
INSERT INTO measurements
|
||||
(id, user_id, type, value, unit, source, measured_at, external_id,
|
||||
sleep_stage, created_at, updated_at)
|
||||
SELECT
|
||||
'seed-' || $1 || '-sleep-' || d || '-' || s,
|
||||
$1, 'SLEEP_DURATION', 16, 'minutes', 'APPLE_HEALTH',
|
||||
((current_date - d)::timestamp - interval '1 hour' + s * interval '16 minutes')
|
||||
AT TIME ZONE 'Europe/Berlin' AT TIME ZONE 'UTC',
|
||||
'hk-' || md5('sleep' || d || '-' || s),
|
||||
(ARRAY['CORE','DEEP','CORE','REM','AWAKE','CORE']::sleep_stage[])[1 + s % 6],
|
||||
now(), now()
|
||||
FROM generate_series(1, $2::int) AS d, generate_series(0, 29) AS s
|
||||
`,
|
||||
USER_ID,
|
||||
days,
|
||||
);
|
||||
total += sleep;
|
||||
console.log(`${"SLEEP_DURATION".padEnd(28)} ${String(sleep).padStart(9)}`);
|
||||
|
||||
await prisma.$executeRawUnsafe(`ANALYZE measurements`);
|
||||
console.log(`${"total".padEnd(28)} ${String(total).padStart(9)}`);
|
||||
}
|
||||
|
||||
main()
|
||||
.catch((err) => {
|
||||
console.error(err);
|
||||
process.exitCode = 1;
|
||||
})
|
||||
.finally(() => prisma.$disconnect());
|
||||
@@ -4,7 +4,7 @@
|
||||
* A self-hoster with 1.25 million measurement rows (a multi-year Apple Health
|
||||
* export) saw the dense intra-day retention, the nightly drain tick and the
|
||||
* data backup all fail with `job timed out`. Reproduced against a seeded
|
||||
* account of the same size (`scripts/seed-large-account.ts`): the drains held
|
||||
* account of the same size (`scripts/bench/seed-large-account.ts`): the drains held
|
||||
* a type's whole history in memory before folding a day, spent about 90
|
||||
* statements per folded day, and paged with a cursor Postgres could not use as
|
||||
* an index condition; the backup added about 640 MB to the process to store
|
||||
|
||||
Reference in new issue
Block a user