Payload template deployed on Railway : memory increasing > 1gb, memory leak ???

[bubbler3]

Hello,
Using almost a payload template with less than 20 pages deployed to Railway with supabase database to store data:
"@payloadcms/next": "3.71.1",
"next": "15.4.10",
I can notice memory consumption goes from 200mb to more than 1gb in less than 3 days while the server being idle with almost no traffic. This causes Railways crashes and additional costs. I dont understand this memory adding up constantly and this seems to be due to a memory leak.

I have already discussed with Railway team trying following solution but with no positive results :
Using Railway variable environment :
NODE_OPTIONS : --max-old-space-size=256 causes then deployment afterward to fail with FatalProcessOutOfMemory

• in package.json:
  start": "cross-env NODE_OPTIONS='--max-old-space-size=300 --expose-gc --no-deprecation' node server.js".

The only temporary solution found is to restart the server every day to maintain the memory under 300mb otherwise it increases quickly.

Have someone met this issue?

[shomriddho]

I don’t think this is a issue with payload. This might be something else using the ram. can you share you payload config and package.json ?

[bubbler3]

Hello,

ok.here my payload config

`// storage-adapter-import-placeholder
import { postgresAdapter } from '@payloadcms/db-postgres'
import { s3Storage } from '@payloadcms/storage-s3'

import sharp from 'sharp' // sharp-import
import path from 'path'
import { buildConfig, LocalizationConfig, PayloadRequest } from 'payload'
import { en } from '@payloadcms/translations/languages/en'
import { fr } from '@payloadcms/translations/languages/fr'
import { fileURLToPath } from 'url'

import { Categories } from './collections/Categories'
import { Media } from './collections/Media'
import { Pages } from './collections/Pages'
import { Posts } from './collections/Posts'
import { Users } from './collections/Users'
import { Footer } from './Footer/config'
import { Header } from './Header/config'
import { plugins } from './plugins'
import { defaultLexical } from '@/fields/defaultLexical'
import { getServerSideURL } from './utilities/getURL'
import localization from './i18n/localization'
import { Topics } from './collections/Topics'


const filename = fileURLToPath(import.meta.url)
const dirname = path.dirname(filename)

export default buildConfig({
  i18n: {
    supportedLanguages: { en, fr },
    fallbackLanguage: 'en',
  },
  localization: localization as LocalizationConfig,
  admin: {
    meta: {
      titleSuffix: '- Blog', // Optionnel
    },
    components: {
      // The `BeforeLogin` component renders a message that you see while logging into your admin panel.
      // Feel free to delete this at any time. Simply remove the line below.
      beforeLogin: ['@/components/BeforeLogin'],
      // The `BeforeDashboard` component renders the 'welcome' block that you see after logging into your admin panel.
      // Feel free to delete this at any time. Simply remove the line below.
      //beforeDashboard: ['@/components/BeforeDashboard'],
      graphics: {
        Logo: '@/components/Logo/AdminLogo#AdminLogo', // Chemin vers votre logo
        Icon: '@/components/Logo/AdminLogo#IconLogo', // Icon pour les onglets
      },
    },
    importMap: {
      baseDir: path.resolve(dirname),
    },
    user: Users.slug,
    livePreview: {
      breakpoints: [
        {
          label: 'Mobile',
          name: 'mobile',
          width: 375,
          height: 667,
        },
        {
          label: 'Tablet',
          name: 'tablet',
          width: 768,
          height: 1024,
        },
        {
          label: 'Desktop',
          name: 'desktop',
          width: 1440,
          height: 900,
        },
      ],
      //url: ({ collectionConfig, data, locale }) => `/${locale}/${collectionConfig?.slug === 'pages' ? (data.slug !== 'home' ? data.slug : '') : ''}`,
    },
  },
  // This config helps us configure global or default features that the other editors can inherit
  editor: defaultLexical,
  db: postgresAdapter({
    pool: {
      connectionString: process.env.DATABASE_URI || '',
      max: 2, // added to reduce memory
      min: 1, // 08/12/2025 added to reduce memory
      idleTimeoutMillis: 10000, // 08/12/2025 added to reduce memory
      connectionTimeoutMillis: 2000,
    },
    // ... your connection pool config
    push: true, // <-- ADD THIS LINE TO DISABLE AUTO PUSHING
  }),
  collections: [Pages, Topics, Posts, Media, Categories, Users],
  cors: [getServerSideURL()].filter(Boolean),
  globals: [Header, Footer],
  plugins: [
    ...plugins,
    // storage-adapter-placeholder
    s3Storage({
      collections: {
        media: {
          prefix: 'media',
        },
      },
      bucket: process.env.S3_BUCKET || '',
      config: {
        forcePathStyle: true, // Important for using Supabase
        credentials: {
          accessKeyId: process.env.S3_ACCESS_KEY_ID || '',
          secretAccessKey: process.env.S3_SECRET_ACCESS_KEY || '',
        },
        region: process.env.S3_REGION,
        endpoint: process.env.S3_ENDPOINT,
      },
    }),
  ],
  secret: process.env.PAYLOAD_SECRET,
  sharp,
  typescript: {
    outputFile: path.resolve(dirname, 'payload-types.ts'),
  },
  jobs: {
    access: {
      run: ({ req }: { req: PayloadRequest }): boolean => {
        // Allow logged in users to execute this endpoint (default)
        if (req.user) return true

        // If there is no logged in user, then check for t
        // for the Vercel Cron secret to be present as an
        // Authorization header:
        const authHeader = req.headers.get('authorization')
        return authHeader === `Bearer ${process.env.CRON_SECRET}`
      },
    },
    tasks: [],
  },
})

`

and package.json :

 `{
  "name": "blog",
  "version": "1.0.0",
  "description": "BubblesMapper Blog",
  "license": "MIT",
  "type": "module",
  "scripts": {
    "build": "node build-with-env.js",
    "dev": "cross-env NODE_OPTIONS=--no-deprecation next dev",
    "dev:prod": "cross-env NODE_OPTIONS=--no-deprecation rm -rf .next && pnpm build && pnpm start",
    "generate:importmap": "cross-env NODE_OPTIONS=--no-deprecation payload generate:importmap",
    "generate:types": "cross-env NODE_OPTIONS=--no-deprecation payload generate:types",
    "ii": "cross-env NODE_OPTIONS=--no-deprecation pnpm --ignore-workspace install",
    "lint": "cross-env NODE_OPTIONS=--no-deprecation next lint",
    "lint:fix": "cross-env NODE_OPTIONS=--no-deprecation next lint --fix",
    "payload": "cross-env NODE_OPTIONS=--no-deprecation payload",
    "reinstall": "cross-env NODE_OPTIONS=--no-deprecation rm -rf node_modules && rm pnpm-lock.yaml && pnpm --ignore-workspace install",
    "start": "cross-env NODE_OPTIONS='--max-old-space-size=300 --expose-gc --no-deprecation' node server.js",
    "test": "pnpm run test:int && pnpm run test:e2e",
    "test:e2e": "cross-env NODE_OPTIONS=\"--no-deprecation --no-experimental-strip-types\" pnpm exec playwright test --config=playwright.config.ts",
    "test:int": "cross-env NODE_OPTIONS=--no-deprecation vitest run --config ./vitest.config.mts"
  },
  "dependencies": {
    "@payloadcms/admin-bar": "3.71.1",
    "@payloadcms/db-postgres": "3.71.1",
    "@payloadcms/live-preview-react": "3.71.1",
    "@payloadcms/next": "3.71.1",
    "@payloadcms/payload-cloud": "3.71.1",
    "@payloadcms/plugin-form-builder": "3.71.1",
    "@payloadcms/plugin-nested-docs": "3.71.1",
    "@payloadcms/plugin-redirects": "3.71.1",
    "@payloadcms/plugin-search": "3.71.1",
    "@payloadcms/plugin-seo": "3.71.1",
    "@payloadcms/richtext-lexical": "3.71.1",
    "@payloadcms/storage-s3": "^3.71.1",
    "@payloadcms/translations": "^3.71.1",
    "@payloadcms/ui": "3.71.1",
    "@radix-ui/react-checkbox": "^1.3.3",
    "@radix-ui/react-label": "^2.1.7",
    "@radix-ui/react-select": "^2.2.6",
    "@radix-ui/react-slot": "^1.2.3",
    "class-variance-authority": "^0.7.1",
    "clsx": "^2.1.1",
    "cross-env": "^7.0.3",
    "dotenv": "16.4.7",
    "geist": "^1.4.2",
    "graphql": "^16.11.0",
    "lucide-react": "^0.378.0",
    "next": "15.4.10",
    "next-intl": "4.3.5",
    "next-sitemap": "^4.2.3",
    "payload": "3.71.1",
    "prism-react-renderer": "^2.4.1",
    "react": "19.1.0",
    "react-dom": "19.1.0",
    "react-hook-form": "7.45.4",
    "sharp": "0.34.2",
    "tailwind-merge": "^2.6.0",
    "tailwindcss-animate": "^1.0.7"
  },
  "devDependencies": {
    "@eslint/eslintrc": "^3.3.1",
    "@playwright/test": "1.54.1",
    "@tailwindcss/typography": "^0.5.16",
    "@testing-library/react": "16.3.0",
    "@types/escape-html": "^1.0.4",
    "@types/node": "22.5.4",
    "@types/react": "19.1.8",
    "@types/react-dom": "19.1.6",
    "@vitejs/plugin-react": "4.5.2",
    "autoprefixer": "^10.4.21",
    "copyfiles": "^2.4.1",
    "eslint": "^9.34.0",
    "eslint-config-next": "15.4.4",
    "jsdom": "26.1.0",
    "playwright": "1.54.1",
    "playwright-core": "1.54.1",
    "postcss": "^8.5.6",
    "prettier": "^3.6.2",
    "tailwindcss": "^3.4.17",
    "typescript": "5.7.3",
    "vite-tsconfig-paths": "5.1.4",
    "vitest": "3.2.3"
  },
  "packageManager": "[email protected]",
  "engines": {
    "node": "^18.20.2 || >=20.9.0",
    "pnpm": "^9 || ^10"
  },
  "pnpm": {
    "onlyBuiltDependencies": [
      "sharp",
      "esbuild",
      "unrs-resolver"
    ]
  }
}`

Thanks for tips

[shomriddho]

Hi @bubbler3 — thanks for sharing the config 👍
Based on this, I still don’t think this is an issue with Payload core itself.

Things to check first

• sharp native memory
  sharp might be the cause of RSS memory growth.
  Try limiting its cache and concurrency:

  import sharp from 'sharp'
  
  sharp.cache({ memory: 50, files: 0, items: 0 })
  sharp.concurrency(1)

  Then monitor RSS vs heap usage.

• Native vs JS heap
  Try Loging memory usage periodically to see what’s actually growing:

  setInterval(() => {
    console.log(process.memoryUsage())
  }, 300000)

  If rss or external keeps increasing while heapUsed stays relatively stable, that usually points to native memory (often sharp or image processing).

• Disable features to isolate
  Try temporarily disabling:

  • s3Storage
  • livePreview
  • plugins
    Then re-enable them one by one until the memory growth returns.

Lowering --max-old-space-size won’t fix a leak and will just cause OOM crashes if the app needs more memory.
Also worth checking whether the same memory growth happens locally — if not, it may be specific to the Railway environment.

[bubbler3]

I have tried to log previously with following server.js code.
In same time, as Railway team also think it is not due to their infrastructure, my current hack is to force the crash and restart every day to flush the memory, otherwise leaks reach 1gb with additional costs.

// server.js

import { createServer } from 'http'
import { parse } from 'url'
import next from 'next'
import { execSync } from 'child_process'

const dev = process.env.NODE_ENV !== 'production'
const hostname = '0.0.0.0'
const port = process.env.PORT || 3000
const app = next({ dev, hostname, port })
const handle = app.getRequestHandler()

let startMemory = process.memoryUsage().rss / 1024 / 1024
let startTime = Date.now()

setInterval(() => {
  try {
    const cgroupMem = execSync(
      'cat /sys/fs/cgroup/memory/memory.usage_in_bytes 2>/dev/null || cat /sys/fs/cgroup/memory.current 2>/dev/null || echo "0"',
    )
      .toString()
      .trim()
    const containerMB = parseInt(cgroupMem) / 1024 / 1024
    const nodeMB = process.memoryUsage().rss / 1024 / 1024

    const uptime = (Date.now() - startTime) / 1000 / 60 / 60
    const growth = nodeMB - startMemory
    const growthPerHour = growth / uptime

    console.log(`🔍 Conteneur : ${containerMB.toFixed(2)} MB`)
    console.log(`📊 Node.js : ${nodeMB.toFixed(2)} MB`)
    console.log(`⚠️ Différence : ${(containerMB - nodeMB).toFixed(2)} MB`)
    console.log(
      `📈 Croissance: +${growth.toFixed(2)} MB en ${uptime.toFixed(1)}h (+${growthPerHour.toFixed(2)} MB/h)`,
    )
  } catch (error) {
    console.error('Erreur:', error.message)
  }
}, 30000)

//  GC SI DISPONIBLE
setInterval(() => {
  if (global.gc) {
    console.log('🧹 Running garbage collection...')
    global.gc()
  }
}, 6 * 60 * 60 * 1000)

// Auto-restart
setTimeout(() => {
  console.log('🔄 Auto-restart after 24h to clear memory')
  process.exit(1)
}, 24 * 60 * 60 * 1000)

app.prepare().then(() => {
  createServer(async (req, res) => {
    try {
      const parsedUrl = parse(req.url, true)
      await handle(req, res, parsedUrl)
    } catch (err) {
      console.error('Error occurred handling', req.url, err)
      res.statusCode = 500
      res.end('internal server error')
    }
  }).listen(port, (err) => {
    if (err) throw err
    console.log(`Server ready on http://${hostname}:${port}`)
    console.log('✅ Memory logger active - checking every 30s')
  })
})

i got :

📈 Croissance: +202.60 MB en 24.0h (+8.45 MB/h)
🔍 Conteneur : 365.44 MB
📊 Node.js : 272.35 MB
⚠️ Différence : 93.09 MB
📈 Croissance: +202.60 MB en 24.0h (+8.44 MB/h)
🔄 Auto-restart after 24h to clear memory
 ELIFECYCLE  Command failed with exit code 1.

[email protected] start /app
cross-env NODE_OPTIONS='--max-old-space-size=300 --expose-gc --no-deprecation' node server.js
✅ Server ready on http://0.0.0.0:8080
✅ Memory logger active - checking every 30s
🔍 Conteneur : 266.46 MB
📊 Node.js : 184.49 MB
⚠️ Différence : 81.96 MB

I will check your tips in payload.config.ts and let you know :
sharp.cache({ memory: 50, files: 0, items: 0 })
sharp.concurrency(1)

[shomriddho]

Thanks for sharing the numbers — this is actually very useful 👍

From your logs:

• Node RSS grows ~8–9 MB/hour while idle
• heapUsed is not shown growing dramatically
• There is a persistent gap between container memory and Node RSS (~80–90 MB)

That strongly suggests this is native / external memory growth, not a JS heap reminder issue. In other words: GC won’t help, and --max-old-space-size won’t fix it either.

This pattern is most likely caused by:

• sharp / libvips native buffers
• image processing pipelines
• or another native dependency retaining memory

A few notes on your current setup:

• The auto-restart confirms it’s not traffic-related — memory resets cleanly after restart
• global.gc() running every 6h not helping is another sign it’s not JS heap
• The container > Node difference is expected when native memory is involved

At this point I’d recommend:

1. Explicitly limiting sharp cache + concurrency (if not already done)
2. Temporarily disabling sharp / media processing entirely to confirm the slope disappears
3. If possible, testing without s3Storage to rule out upload/image pipelines
4. Avoid relying on forced restarts long-term — good as a mitigation, but not a fix

Given the steady linear growth while idle, this still doesn’t look like Payload core itself, but rather a native dependency behaving poorly in a long-lived process.

If disabling sharp stops the growth, that’s almost certainly the root cause.

[bubbler3]

Thank you for detailed explanation. It helped me to dig around this issue.
Instructions as

sharp.cache({ memory: 50, files: 0, items: 0 })
sharp.concurrency(1)

didnt work so well as the log still show a steaty growth

🔍 Conteneur : 444.34 MB
📊 Node.js : 406.39 MB
⚠️ Différence : 37.95 MB
📈 Croissance: +335.32 MB en 14.7h (+22.80 MB/h)

I understand that this issue with sharp / libvips is already documented. After images processing , memory is given back to the system allocator but not to the OS, counting still into the RSS, which is the one on which services as Railway are invoicing.

I am testing 'sharp removal ' from Payload config currently but doing this stopped me from having the image processing/sizing enabled by sharp into payload. I need to wait a couple of days to confirm but already without sharp, memory went back to 200mb.

As such behaviour of retaining memory is not considered a sharp bug and that only a restart can destroy the process and free the memory, this should be documented as a drawback of using Payload+Railway stack.

I understand force Railway Auto-restart doesnt seem a long term solution but if i keep Sharp library, i didnt find any other command to ask Railway to restart. Do you know one?

setTimeout(() => {
  console.log('🔄 Auto-restart after 24h to clear memory')
  process.exit(1)
}, 24 * 60 * 60 * 1000) 

[bubbler3]

ok, after testing without sharp package, the memory in production still progressively increased up from 200mb 800mb after 4 days, while the blog has almost no traffic and a couple of pages.
Thanks @shomriddho for looking into that matter, but Payload with Railway is still a matter.
As i use a payload template, i hope in near future to get some feedback from other dev using the same stack.

📊 Node.js : 576.09 MB
⚠️ Différence : 120.45 MB
📈 Croissance: +500.80 MB en 97.6h (+5.13 MB/h)

so my only current adopted solution is to agressively restart every day with command :

 setTimeout(() => {
  console.log('🔄 Auto-restart after 24h to clear memory')
  process.exit(1)
}, 24 * 60 * 60 * 1000)