You push your code to Hostinger, reload your domain, and get slapped with a blank screen and a “503 Service Unavailable” error. It’s frustrating, but it almost always means the exact same thing: your Node process crashed before it ever handled its first request. We’re going to dig into the runtime, pull the actual stderr logs over SSH, and fix whatever killed it.
Hostinger doesn’t run Node apps like a typical VPS where you set up a custom systemd unit. On their shared and Cloud plans, Node runs behind OpenLiteSpeed or Nginx managed by CloudLinux and Phusion Passenger. When Passenger tries to boot your entry file and the process exits with an error, the upstream reverse proxy gives up waiting and throws a 503 Service Unavailable back to the browser.

Why Hostinger Returns 503 on Node.js Apps
Node didn’t generate that 503 error—the edge web server did. The proxy expected your app to stay running on an internal socket or port, and instead it found nothing listening.
In practice, 95% of Node 503s on Hostinger trace back to four culprits:
- Uncaught boot-time exceptions: A missing environment variable, bad import, or syntax error that crashes your script the second it starts.
- Hardcoded ports: Binding directly to port 3000 instead of listening on the port or socket Passenger provides.
- Missing or mismatched dependencies: Forgetting to run a build, or copying native binary modules across from your local machine.
- CloudLinux LVE memory limits: Exceeding your plan’s strict RAM cap, causing the OS kernel to silently kill your process.
If you followed our walkthrough to deploy a Node.js app on Hostinger, you’re using this standard CloudLinux Passenger setup. Here is how I track down what went wrong.
Step 1: Check the Real Application Error Logs
Stop refreshing your browser hoping it’ll magically resolve. You need the actual stack trace from Passenger’s stderr output.
SSH into your Hostinger account or open the terminal inside hPanel. Depending on your specific hosting plan, error logs usually end up in one of two spots: your project’s root folder or the LiteSpeed log directory.
First, run this inside your app’s directory to see if Passenger dumped boot logs there:
# Navigate to your application root
cd ~/domains/yourdomain.com/public_html # Check if passenger wrote a crash log
ls -la passenger.*.log # Tail the most recent error log
tail -n 50 passenger.*.logIf that file doesn’t exist, check the stderr stream that Node writes directly into your virtual environment directory:
# Check the node virtual environment logs
cat ~/nodevenv/domains/yourdomain.com/public_html/18/lib/stderr.logSwap out 18 for whichever Node version you picked in hPanel (18, 20, etc.). If your script threw an uncaught error at boot, the stack trace is sitting right here—file name, line number, and all.
Step 2: Fix Entry Point and Application Startup Path Mismatches
Hostinger requires you to point directly to your startup file inside hPanel under Websites > Manage > Node.js. If hPanel expects app.js, but your build outputs to dist/server.js or your root file is called index.js, Passenger crashes immediately and serves a 503.
Make sure your package.json matches your hPanel setup. Open your package.json file:
{ "name": "my-backend-app", "version": "1.0.0", "main": "app.js", "scripts": { "start": "node app.js" }, "dependencies": { "express": "^4.19.2", "dotenv": "^16.4.5" }
}Then check these three settings in the hPanel Node.js dashboard:
- Application root: Points to your project folder (e.g.,
domains/yourdomain.com/public_html, or a subfolder if you run a monorepo). - Application startup file: The exact entry file name (like
app.jsorindex.js). Don’t prefix it with./. - Node.js version: Needs to support whatever syntax your code uses. If you rely on native fetch or top-level await on an older runtime, the script will crash on line one.
Step 3: Stop Hardcoding Local Ports
This catches almost everyone deploying to shared hosting for the first time. On your local machine, you probably run app.listen(3000). On Hostinger, Passenger manages incoming connections dynamically. It routes traffic through a custom port or Unix socket via the PORT environment variable.
If you hardcode port 3000, your app might technically boot in the background, but the reverse proxy has no idea how to talk to it. After roughly 30 seconds, the proxy times out and drops a 503.
Open your entry file (like app.js or server.js) and bind to the dynamic port:
const express = require('express');
const app = express(); // Use the port assigned by Hostinger, or fallback for local dev
const PORT = process.env.PORT || 3000; app.get('/', (req, res) => { res.status(200).send('API is healthy and running.');
}); // Bind to 0.0.0.0 so the internal reverse proxy can forward traffic
const server = app.listen(PORT, '0.0.0.0', () => { console.log(`Server listening on port ${PORT}`);
}); // Handle termination signals gracefully
process.on('SIGTERM', () => { server.close(() => { console.log('Process terminated cleanly'); });
});For more details on passing runtime secrets and managing ports safely, check out our guide on how to configure environment variables in Node.js on Hostinger.
Step 4: Resolve Native Module and Architecture Mismatches
Did you upload your local node_modules folder straight from a Mac or Windows machine using SFTP or Git? If you did, that’s almost certainly your problem.
Packages with native C++ bindings—like bcrypt, sharp, or better-sqlite3—compile binaries for the OS you ran npm install on. If you drop macOS ARM64 binaries onto Hostinger’s Linux x86_64 servers, Node will throw an invalid ELF header error and die instantly.
Wipe the remote dependencies and rebuild them directly on Hostinger’s environment:
# Log into SSH and enter the project folder
cd ~/domains/yourdomain.com/public_html # Remove existing local-uploaded dependencies
rm -rf node_modules package-lock.json # Clean the npm cache
npm cache clean --force # Reinstall all modules natively on the Linux host
npm install --productionIf you’re deploying a modern SSR framework like Nuxt, take a look at our guide on how to deploy Nuxt 3 on Hostinger so your Nitro build outputs clean, Linux-ready production assets.
Step 5: Inspect CloudLinux LVE Limits and Out-of-Memory Crashes
Hostinger isolates shared accounts using CloudLinux LVE. If your plan gives you 512 MB or 1 GB of RAM and your Node process tries to grab more during startup or heavy traffic, the kernel will kill it instantly (OOM). When that happens, you won’t even get a clean JavaScript stack trace.
To verify if the OS killed your app, go to hPanel under Hosting > Resource Usage. Look for sudden spikes in the RAM or Faults graphs that line up with when the 503 happened.
You can also check your current memory allocations over SSH:
# Check total and used system memory
free -m # Check process limits applied to your user
ulimit -aBy default, Node’s V8 engine can try to reserve over 1 GB of heap space, which will instantly choke small shared hosting tiers. You can force it to stay within boundaries by passing a max heap limit in your startup arguments:
# Start node with a strict 256MB heap limit
node --max-old-space-size=256 app.jsAdd this flag to your package.json start script or paste it into the custom arguments box in the hPanel Node.js manager.
Step 6: Trigger a Clean Passenger App Restart
Pushing files over SFTP or Git doesn’t always make Passenger reload your Node app. More often than not, the old crashed process stays cached in memory or gets caught in a restart loop. You need to force an explicit restart.
Passenger watches for changes to a file called restart.txt inside your tmp directory. You can trigger an instant reload over SSH with this:
# Ensure the tmp folder exists inside your application root
mkdir -p ~/domains/yourdomain.com/public_html/tmp # Touch the restart file to trigger a reload
touch ~/domains/yourdomain.com/public_html/tmp/restart.txtYou can also hit the Restart button directly inside the hPanel Node.js dashboard. Give it a few seconds, then test your site with curl to see the raw HTTP status:
# Inspect HTTP status and response headers
curl -I https://yourdomain.comIf you sorted out the underlying crash, you’ll see a clean 200 OK instead of the 503 screen.
Frequently Asked Questions
Can I use PM2 to manage my Node.js app on Hostinger shared plans?
No. CloudLinux and Passenger manage the processes here. If you try running a background daemon like PM2, Hostinger’s process reaper will kill it as soon as your SSH session disconnects. You have to let Passenger supervise the app.
Why does my app work locally on port 3000 but fail on Hostinger?
Locally, your browser connects directly to Node’s internal web server on 3000. On Hostinger, users connect to OpenLiteSpeed or Nginx first. That server then proxies requests to your app using a dynamic socket or port injected through process.env.PORT. If your code ignores that variable, the proxy can’t forward requests, causing a 503.
How do I run database migrations without causing a 503 timeout?
Don’t bundle heavy migration commands (like Prisma or TypeORM migrations) into your main startup script. If boot takes longer than 30 to 60 seconds, Passenger assumes your process hung and cuts it off with a 503. Run your migrations manually in the SSH terminal before restarting the app.
Where are the official docs for troubleshooting Passenger Node apps?
Check the Phusion Passenger Library and the Node.js HTTP documentation for details on socket bindings, process signals, and lifecycle management.
Next Steps for Your Deployment
Now that your application is answering with 200s, make sure your sensitive database credentials and tokens aren’t exposed in your code repo. Check out our walkthrough on configuring environment variables in Node.js on Hostinger to lock down your production config.

