feat(ops): health-fail alerts, optional DB backup, admin UX polish

Wire jobs-worker health probes to Discord/email alerts with cooldown, optional mysqldump, rate-limit /api/health, mark-all-read alerts, ConfirmDialog on destructive admin actions, and raise coverage floors.

Co-authored-by: Cursor <[email protected]>
This commit is contained in:
SimoandCursor committed 2026-08-01 15:25:47 +02:00
1 parent 3bd712e744
commit 725e1cb338
16 files changed
+327 -32

No files matched your search

+160 -1
View File
@@ -1,9 +1,12 @@
import * as Sentry from "@sentry/nextjs";
import { Cron } from "croner";
import { lt } from "drizzle-orm";
import { lt, sql } from "drizzle-orm";
import { env } from "../src/env";
import { db, PasswordReset, WebsiteLoginLogs } from "../src/lib/db";
import { logger } from "../src/lib/logger";
import { redis } from "../src/lib/redis";
import { emulatorOffline, healthDegraded } from "../src/lib/services/alert";
import { rcon } from "../src/lib/services/rcon";
function initWorkerSentry(): void {
const dsn = process.env.SENTRY_DSN;
@@ -28,6 +31,71 @@ function captureWorkerError(err: unknown, context: string): void {
}
}
/** In-process cooldown so a flapping probe does not spam Discord/email. */
const alertCooldownMs = (env.HEALTH_ALERT_COOLDOWN_MIN ?? 15) * 60_000;
const lastHealthAlertAt = new Map<string, number>();
function canAlert(key: string): boolean {
const now = Date.now();
const prev = lastHealthAlertAt.get(key) ?? 0;
if (now - prev < alertCooldownMs) return false;
lastHealthAlertAt.set(key, now);
return true;
}
async function probeHealth(): Promise<{
database: boolean;
redis: boolean | null;
emulator: boolean;
}> {
const database = await db
.execute(sql`SELECT 1`)
.then(() => true)
.catch(() => false);
let redisOk: boolean | null = null;
if (env.REDIS_URL) {
if (!redis) {
redisOk = false;
} else {
try {
redisOk = (await redis.ping()) === "PONG";
} catch {
redisOk = false;
}
}
}
const emulator = await rcon.send("ping", null).catch(() => false);
return {
database,
redis: redisOk,
emulator: Boolean(emulator),
};
}
async function checkOpsHealth(): Promise<void> {
try {
const health = await probeHealth();
const degraded =
!health.database || health.redis === false || !health.emulator;
if (!degraded) return;
if (!health.emulator && health.database && health.redis !== false) {
if (canAlert("emulator")) {
await emulatorOffline("jobs-worker RCON ping failed");
}
return;
}
if (canAlert("health")) {
await healthDegraded(health);
}
} catch (err) {
captureWorkerError(err, "Health probe failed");
}
}
async function backupEmulatorJar(): Promise<void> {
if (!env.EMULATOR_JAR_PATH || !env.EMULATOR_BACKUP_DIR) return;
@@ -70,6 +138,82 @@ async function backupEmulatorJar(): Promise<void> {
}
}
/** Optional mysqldump when DB_BACKUP_DIR is set (host must have mysqldump on PATH). */
async function backupDatabase(): Promise<void> {
if (!env.DB_BACKUP_DIR || !env.DATABASE_URL) return;
const { mkdirSync, readdirSync, unlinkSync, existsSync, createWriteStream } =
await import("node:fs");
const { resolve } = await import("node:path");
const { spawn } = await import("node:child_process");
let parsed: URL;
try {
parsed = new URL(env.DATABASE_URL);
} catch {
logger.error("Invalid DATABASE_URL for DB backup", { module: "jobs" });
return;
}
if (!existsSync(env.DB_BACKUP_DIR)) {
mkdirSync(env.DB_BACKUP_DIR, { recursive: true });
}
const timestamp = new Date().toISOString().slice(0, 19).replace(/[T:]/g, "-");
const dbName =
decodeURIComponent(parsed.pathname.replace(/^\//, "")) || "cms";
const outFile = resolve(env.DB_BACKUP_DIR, `db-${dbName}-${timestamp}.sql`);
const args = [
`-h${parsed.hostname}`,
`-P${parsed.port || "3306"}`,
`-u${decodeURIComponent(parsed.username)}`,
`--single-transaction`,
`--routines`,
`--databases`,
dbName,
];
if (parsed.password) {
args.splice(3, 0, `-p${decodeURIComponent(parsed.password)}`);
}
await new Promise<void>((resolvePromise) => {
const child = spawn("mysqldump", args, {
stdio: ["ignore", "pipe", "pipe"],
});
const out = createWriteStream(outFile);
child.stdout.pipe(out);
let stderr = "";
child.stderr.on("data", (chunk: Buffer) => {
stderr += chunk.toString();
});
child.on("error", (err) => {
captureWorkerError(err, "mysqldump spawn failed (is it on PATH?)");
resolvePromise();
});
child.on("close", (code) => {
out.end();
if (code !== 0) {
captureWorkerError(
new Error(stderr || `mysqldump exit ${code}`),
"DB backup failed",
);
} else {
logger.info("Backed up database", { module: "jobs", outFile });
const keep = env.DB_BACKUP_KEEP ?? 7;
const files = readdirSync(env.DB_BACKUP_DIR)
.filter((f) => f.startsWith("db-") && f.endsWith(".sql"))
.sort()
.reverse();
for (let i = keep; i < files.length; i++) {
unlinkSync(resolve(env.DB_BACKUP_DIR, files[i]));
}
}
resolvePromise();
});
});
}
async function cleanupOldLogs(): Promise<void> {
try {
const cutoff = new Date(Date.now() - 30 * 24 * 60 * 60 * 1000);
@@ -107,6 +251,15 @@ async function main() {
});
}
if (env.DB_BACKUP_DIR) {
new Cron("30 3 * * *", () => {
backupDatabase().catch((e) => captureWorkerError(e, "DB backup error"));
});
logger.info("Scheduled: mysqldump DB backup (daily 03:30)", {
module: "jobs",
});
}
new Cron("0 4 * * *", () => {
Promise.all([cleanupOldLogs(), cleanupOldSessions()]).catch((e) =>
captureWorkerError(e, "Cleanup error"),
@@ -114,10 +267,16 @@ async function main() {
});
logger.info("Scheduled: old data cleanup (daily 04:00)", { module: "jobs" });
new Cron("*/5 * * * *", () => {
checkOpsHealth().catch((e) => captureWorkerError(e, "Health check error"));
});
logger.info("Scheduled: ops health probe (every 5 min)", { module: "jobs" });
await Promise.all([
backupEmulatorJar(),
cleanupOldLogs(),
cleanupOldSessions(),
checkOpsHealth(),
]);
}