fix

.gitignore

@@ -1,5 +1,7 @@
 node_modules
 dist
+dist-next
+dist-previous
 .env
 .env.local
 .DS_Store
@@ -7,4 +9,4 @@ npm-debug.log*
 vite-dev*.log
 server-local*.log
 .local-storage/
-.claude/
+.claude/

README.md

@@ -34,7 +34,12 @@ npm run build
 pm2 start ecosystem.config.cjs
 ```
 
-The production server runs on <http://localhost:3010>. It serves `/health`
+The production server runs on <http://localhost:3010>. PM2 starts the web app in
+cluster mode with two workers by default, waits for each worker to signal that it
+is ready, and then reloads workers one at a time during deploys. Override the
+worker count with `WEB_INSTANCES`.
+
+The server serves `/health`
 locally and only proxies the API routes used by the app:
 
 - `GET /api/tss/leaderboard/teams?limit=1..100`
@@ -171,14 +176,17 @@ The default deploy flow is:
 
 ```sh
 git pull --ff-only
-npm install --production=false --include=dev --include=optional
+npm ci --include=dev --include=optional
-npm run build
+npm run build -- --outDir dist-next
+# the webhook promotes dist-next to dist after carrying over old hashed assets
 pm2 reload tssbot-web --update-env
 ```
 
 Only processes listed in `PM2_RESTART_TARGETS` are reloaded. The default is
-`tssbot-web`, so unrelated PM2 processes are left alone. The webhook exits after
+`tssbot-web`, so unrelated PM2 processes are left alone. The web server handles
-24 hours so PM2 restarts it cleanly.
+`SIGINT` and `SIGTERM` by closing its listener and SQLite handles before exit,
+which lets PM2 finish reloads without dropping active requests. The webhook
+exits after 24 hours so PM2 restarts it cleanly.
 
 When webhook code changes are deployed, restart the webhook process once so PM2
 loads the updated listener:

ecosystem.config.cjs

@@ -4,6 +4,11 @@ module.exports = {
       name: 'tssbot-web',
       script: 'server.cjs',
       cwd: __dirname,
+      exec_mode: 'cluster',
+      instances: process.env.WEB_INSTANCES || 2,
+      wait_ready: true,
+      listen_timeout: 10000,
+      kill_timeout: 10000,
       env: {
         NODE_ENV: 'production',
         PORT: process.env.PORT || 3010,

server.cjs

@@ -59,6 +59,7 @@ const MAX_TEAM_NAME_LENGTH = 80
 const MAX_CACHE_ENTRIES = 200
 const MAX_RATE_LIMIT_KEYS = 1000
 const MAX_ANALYTICS_BODY_BYTES = 16 * 1024
+const RUN_BACKGROUND_JOBS = !process.env.NODE_APP_INSTANCE || process.env.NODE_APP_INSTANCE === '0'
 
 const TRUST_PROXY = (() => {
   const raw = String(process.env.TRUST_PROXY ?? 'cloudflare').trim().toLowerCase()
@@ -241,6 +242,7 @@ function ensureAnalyticsDb() {
 
   analyticsDb = new Database(path.join(storageDir, ANALYTICS_DATABASE_FILE))
   analyticsDb.pragma('journal_mode = WAL')
+  analyticsDb.pragma('busy_timeout = 5000')
   analyticsDb.exec(`
     create table if not exists viewer_events (
       id integer primary key autoincrement,
@@ -331,6 +333,7 @@ function ensureUptimeDb() {
 
   uptimeDb = new Database(path.join(storageDir, UPTIME_DATABASE_FILE))
   uptimeDb.pragma('journal_mode = WAL')
+  uptimeDb.pragma('busy_timeout = 5000')
   uptimeDb.exec(`
     create table if not exists uptime_snapshots (
       id integer primary key autoincrement,
@@ -433,12 +436,14 @@ async function uptimeHistory() {
   }
 }
 
+let uptimeSamplerTimer = null
+
 function startUptimeSampler() {
   takeUptimeSnapshot().catch((error) => {
     console.error('Initial uptime snapshot failed:', error)
   })
 
-  setInterval(() => {
+  uptimeSamplerTimer = setInterval(() => {
     takeUptimeSnapshot().catch((error) => {
       console.error('Uptime snapshot failed:', error)
     })
@@ -926,15 +931,17 @@ function purgeOldAnalytics(db) {
   const eventCutoff = new Date(Date.now() - ANALYTICS_RETENTION_DAYS * 24 * 60 * 60 * 1000).toISOString()
   const activeCutoff = new Date(Date.now() - ANALYTICS_ACTIVE_WINDOW_SECONDS * 3 * 1000).toISOString()
 
-  db.prepare(`
+  db.transaction(() => {
-    delete from viewer_events
+    db.prepare(`
-     where occurred_at < ?
+      delete from viewer_events
-  `).run(eventCutoff)
+       where occurred_at < ?
+    `).run(eventCutoff)
 
-  db.prepare(`
+    db.prepare(`
-    delete from active_viewers
+      delete from active_viewers
-     where last_seen_at < ?
+       where last_seen_at < ?
-  `).run(activeCutoff)
+    `).run(activeCutoff)
+  })()
 }
 
 function recordViewerEvent(req, payload) {
@@ -978,44 +985,48 @@ function recordViewerEvent(req, payload) {
   }
 
   const now = new Date().toISOString()
-  db.prepare(`
+  const writeViewerEvent = db.transaction(() => {
-    insert into viewer_events
+    db.prepare(`
-      (occurred_at, visitor_id, session_id, ip_hash, event_type, page_path, page_title,
+      insert into viewer_events
-       referrer, user_agent, browser, os, device, screen, language, timezone,
+        (occurred_at, visitor_id, session_id, ip_hash, event_type, page_path, page_title,
-       country, region, city, latitude, longitude, consent, metadata)
+         referrer, user_agent, browser, os, device, screen, language, timezone,
-    values
+         country, region, city, latitude, longitude, consent, metadata)
-      (@occurred_at, @visitor_id, @session_id, @ip_hash, @event_type, @page_path, @page_title,
+      values
-       @referrer, @user_agent, @browser, @os, @device, @screen, @language, @timezone,
+        (@occurred_at, @visitor_id, @session_id, @ip_hash, @event_type, @page_path, @page_title,
-       @country, @region, @city, @latitude, @longitude, @consent, @metadata)
+         @referrer, @user_agent, @browser, @os, @device, @screen, @language, @timezone,
-  `).run({ ...event, occurred_at: now })
+         @country, @region, @city, @latitude, @longitude, @consent, @metadata)
+    `).run({ ...event, occurred_at: now })
 
-  db.prepare(`
+    db.prepare(`
-    insert into active_viewers
+      insert into active_viewers
-      (session_id, visitor_id, ip_hash, first_seen_at, last_seen_at, page_path, page_title,
+        (session_id, visitor_id, ip_hash, first_seen_at, last_seen_at, page_path, page_title,
-       referrer, user_agent, browser, os, device, screen, language, timezone,
+         referrer, user_agent, browser, os, device, screen, language, timezone,
-       country, region, city, latitude, longitude)
+         country, region, city, latitude, longitude)
-    values
+      values
-      (@session_id, @visitor_id, @ip_hash, @now, @now, @page_path, @page_title,
+        (@session_id, @visitor_id, @ip_hash, @now, @now, @page_path, @page_title,
-       @referrer, @user_agent, @browser, @os, @device, @screen, @language, @timezone,
+         @referrer, @user_agent, @browser, @os, @device, @screen, @language, @timezone,
-       @country, @region, @city, @latitude, @longitude)
+         @country, @region, @city, @latitude, @longitude)
-    on conflict(session_id) do update set
+      on conflict(session_id) do update set
-      last_seen_at = excluded.last_seen_at,
+        last_seen_at = excluded.last_seen_at,
-      page_path = excluded.page_path,
+        page_path = excluded.page_path,
-      page_title = excluded.page_title,
+        page_title = excluded.page_title,
-      referrer = excluded.referrer,
+        referrer = excluded.referrer,
-      user_agent = excluded.user_agent,
+        user_agent = excluded.user_agent,
-      browser = excluded.browser,
+        browser = excluded.browser,
-      os = excluded.os,
+        os = excluded.os,
-      device = excluded.device,
+        device = excluded.device,
-      screen = excluded.screen,
+        screen = excluded.screen,
-      language = excluded.language,
+        language = excluded.language,
-      timezone = excluded.timezone,
+        timezone = excluded.timezone,
-      country = excluded.country,
+        country = excluded.country,
-      region = excluded.region,
+        region = excluded.region,
-      city = excluded.city,
+        city = excluded.city,
-      latitude = excluded.latitude,
+        latitude = excluded.latitude,
-      longitude = excluded.longitude
+        longitude = excluded.longitude
-  `).run({ ...event, now })
+    `).run({ ...event, now })
+  })
+
+  writeViewerEvent()
 }
 
 function deleteViewerData(payload) {
@@ -1790,11 +1801,49 @@ const server = http.createServer((req, res) => {
 server.listen(PORT, '0.0.0.0', () => {
   console.log(`tssbot-web serving http://localhost:${PORT}`)
   console.log(`proxying API requests to ${API_UPSTREAM}`)
-  console.log(`sampling uptime every ${Math.round(UPTIME_SAMPLE_INTERVAL_MS / 60000)} minutes`)
+  if (RUN_BACKGROUND_JOBS) {
+    console.log(`sampling uptime every ${Math.round(UPTIME_SAMPLE_INTERVAL_MS / 60000)} minutes`)
+  } else {
+    console.log('uptime sampler disabled in this worker')
+  }
   console.log(`storing uptime snapshots in ${path.join(uptimeStoragePath(), UPTIME_DATABASE_FILE)}`)
   console.log(`storing viewer analytics in ${path.join(uptimeStoragePath(), ANALYTICS_DATABASE_FILE)}`)
   if (!TURNSTILE_SECRET_KEY) {
     console.warn('TURNSTILE_SECRET_KEY is not set — Turnstile verification is disabled and gated endpoints will accept any request')
   }
-  startUptimeSampler()
+  if (RUN_BACKGROUND_JOBS) startUptimeSampler()
+  process.send?.('ready')
 })
+
+let shuttingDown = false
+
+function closeDatabase(db, name) {
+  if (!db) return
+
+  try {
+    db.close()
+  } catch (error) {
+    console.error(`Failed to close ${name} database:`, error)
+  }
+}
+
+function shutdown() {
+  if (shuttingDown) return
+  shuttingDown = true
+
+  if (uptimeSamplerTimer) clearInterval(uptimeSamplerTimer)
+
+  server.close(() => {
+    closeDatabase(uptimeDb, 'uptime')
+    closeDatabase(analyticsDb, 'analytics')
+    process.exit(0)
+  })
+
+  setTimeout(() => {
+    console.error('Graceful shutdown timed out')
+    process.exit(1)
+  }, 10000).unref()
+}
+
+process.on('SIGINT', shutdown)
+process.on('SIGTERM', shutdown)

webhook.cjs

@@ -44,6 +44,9 @@ const RESTART_TARGETS = (process.env.PM2_RESTART_TARGETS || 'tssbot-web')
   .split(',')
   .map((target) => target.trim())
   .filter(Boolean)
+const DIST_DIR = path.join(__dirname, 'dist')
+const NEXT_DIST_DIR = path.join(__dirname, 'dist-next')
+const PREVIOUS_DIST_DIR = path.join(__dirname, 'dist-previous')
 const ALLOWED_REFS = new Set(
   (process.env.GITHUB_WEBHOOK_REFS || 'refs/heads/main')
     .split(',')
@@ -198,6 +201,49 @@ async function ensureBuildDependencies() {
   }
 }
 
+function copyMissingFiles(fromDir, toDir) {
+  if (!fs.existsSync(fromDir) || !fs.existsSync(toDir)) return
+
+  for (const entry of fs.readdirSync(fromDir, { withFileTypes: true })) {
+    const source = path.join(fromDir, entry.name)
+    const target = path.join(toDir, entry.name)
+
+    if (entry.isDirectory()) {
+      fs.mkdirSync(target, { recursive: true })
+      copyMissingFiles(source, target)
+      continue
+    }
+
+    if (!fs.existsSync(target)) {
+      fs.copyFileSync(source, target)
+    }
+  }
+}
+
+function promoteBuiltDist() {
+  const previousAssetsDir = path.join(DIST_DIR, 'assets')
+  const nextAssetsDir = path.join(NEXT_DIST_DIR, 'assets')
+  let movedCurrentDist = false
+
+  copyMissingFiles(previousAssetsDir, nextAssetsDir)
+
+  fs.rmSync(PREVIOUS_DIST_DIR, { recursive: true, force: true })
+
+  try {
+    if (fs.existsSync(DIST_DIR)) {
+      fs.renameSync(DIST_DIR, PREVIOUS_DIST_DIR)
+      movedCurrentDist = true
+    }
+
+    fs.renameSync(NEXT_DIST_DIR, DIST_DIR)
+  } catch (error) {
+    if (movedCurrentDist && !fs.existsSync(DIST_DIR) && fs.existsSync(PREVIOUS_DIST_DIR)) {
+      fs.renameSync(PREVIOUS_DIST_DIR, DIST_DIR)
+    }
+    throw error
+  }
+}
+
 function postDiscordWebhook(payload) {
   if (!DISCORD_WEBHOOK_URL) return Promise.resolve()
 
@@ -360,7 +406,8 @@ async function deploy(push) {
     await run('git', ['pull', '--ff-only'])
     diff = await deployDiff(push)
     await ensureBuildDependencies()
-    await run('npm', ['run', 'build'])
+    await run('npm', ['run', 'build', '--', '--outDir', 'dist-next'])
+    promoteBuiltDist()
 
     for (const target of RESTART_TARGETS) {
       await run('pm2', ['reload', target, '--update-env'])