Github Actions & EC2 Deployment: Troubleshooting Guide

• Issue 1: 502 Bad Request
  • Understanding the Issue
  • Fix
• Issue 2: Failed Authentication and Data Fetching
  • Understanding the Issue
  • Fix
• Issue 3: Running Out of Memory
  • Understanding the Issue
  • Fix
• Issue 4: Running Out of Disk Space
  • Understanding the Issue
  • Fix
• Conclusion

When working on Assignment 1 of the Software Life Cycle Management unit, we ran into several technical issues, even though we followed the instructions provided by the teaching staff.

Fortunately, by understanding the role of each step, I was able to troubleshoot the problems and get my application running stably on AWS EC2. In this article, I'll share some of the most common issues and how I fixed them.

Issue 1: 502 Bad Request

Understanding the Issue

The error means Nginx cannot forward your request to the application server (in our case, the Node.js server). The most likely reason is that the backend process isn't running.

Fix

Check if your backend process is running with:

pm2 status

If it shows the process as stopped or errored, or if no process is listed at all, start it again from the backend directory:

pm2 start "npm run start" --name="backend"

If that doesn’t solve it, another possible cause is that your GitHub Actions runner is still running. Check the Actions tab on GitHub and wait until all jobs finish before trying the pm2 command again.

Issue 2: Failed Authentication and Data Fetching

Understanding the Issue

If you can access your app from your EC2 instance's public IP, but cannot authenticate or fetch data, the most likely cause is that the base URL in your codebase doesn't match your EC2 public IP. Since the public IP changes each time the instance is restarted, you need to update your base URL accordingly.

Fix

Check if the base URL for data fetching matches your current EC2 public IP. If not, update it in your codebase and push the changes in your main branch.

Issue 3: Running Out of Memory

Understanding the Issue

During CI/CD, I repeatedly got stuck at the yarn run build step. Specifically, the job stopped at:

Creating an optimized production build...

Then it failed with a message that resources weren’t sufficient and the process was killed. This indicates RAM ran out, so you need more memory for the job to succeed.

Fix

You can add swap space, which is a portion of your EC2 disk that acts as overflow memory when RAM is full.

Run the following commands to create extra memory space:

# Check memory usage (h stands for human-readable format)
free -h

# Allocate 2GB disk space
sudo fallocate -l 2G /swapfile

# Grant read and write permissions to file owner
sudo chmod 600 /swapfile

# Make swapfile
sudo mkswap /swapfile

# Enable swap
sudo swapon /swapfile

# Enable swap automatically when rebooting
echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab

Re-run the failed jobs after enabling swap. This method is quick, free, and effective for preventing yarn run build from crashing.

That said, the root cause is the limited performance of the small EC2 instance (t2.micro) used for the assignment. In real-world projects, it’s better to choose a larger, more powerful instance type.

Issue 4: Running Out of Disk Space

Understanding the Issue

After enabling swap space, you may still get stuck again at:

Creating an optimized production build...

But this time, the error is different:

Failed to compile.
ENOSPC: no space left on device, write

This means the disk itself is full, so you need to free up space.

Fix

Some ways to reclaim space include removing packages, clearing caches, and deleting old logs. The following commands worked for me:

# Clear yarn cache
yarn cache clean

# Remove frontend packages
rm -rf node_modules

# Clear old Linux packages
sudo apt-get clean

# Delete system logs
sudo journalctl --vacuum-time=3d

# Clear package cache
sudo rm -rf /var/cache/apt/archives/*

Once cleaned, re-run your failed GitHub Actions jobs.

Conclusion

In this article, I shared the most common issues my classmates and I encountered when implementing CI/CD with GitHub Actions and AWS EC2. By understanding the causes behind the errors, I not only managed to fix them myself but also helped friends troubleshoot their deployments.

Looking back, most of the challenges were due to the limitations of the t2.micro EC2 instance we used for the assignment. In real-world projects, opting for a more capable instance type would eliminate most of these problems entirely.