PM2: The Process Manager for Production Node.js

Why it exists

Running node app.js directly in a terminal is fragile. If the application crashes, it stays down. If you close the terminal, the process dies. For a production application that needs to be available 24/7, you need a separate system to manage the application's process and ensure it remains online.

The mental model

PM2 is a daemon process manager. Think of it as a dedicated supervisor for your application processes. Instead of you directly running node app.js, you tell PM2 to run it. PM2 then starts your app in the background, watches over it, and automatically restarts it if it crashes or runs out of memory, ensuring high availability.

How it works

You install PM2 globally via npm and use its command-line interface. The command pm2 start app.js starts your application and daemonizes it, meaning it runs in the background, detached from your terminal. PM2 maintains a process list you can view with pm2 list. You can then manage your apps using commands like pm2 restart, pm2 stop, and pm2 delete. For persistence across server reboots, you run pm2 startup once to generate a system startup script, then pm2 save anytime you want to freeze the current list of running applications for automatic respawning.

When to use it

Use PM2 for any Node.js application in a production or staging environment that needs to be long-running, like web servers, APIs, or background workers. Its cluster mode (pm2 start app.js -i max) is invaluable for instantly scaling a stateless app to use all available CPU cores, improving performance without changing application code. The --watch flag is also useful in development for automatically restarting the server on file changes.

When not to use it

PM2 is often overkill for simple, one-off scripts. In modern containerized environments like Docker and Kubernetes, its role is diminished. While you can run PM2 inside a container, the container orchestrator (like Kubernetes) is typically responsible for health checks, scaling, and restarts, making PM2's core supervision features redundant. In those cases, running the node process directly is often simpler.

One canonical example

A common use case is deploying a Node.js API. After deploying your code to a server, you would run pm2 start server.js --name my-api -i max. This command starts the application, names it "my-api" for easy management, and uses the -i max flag to launch a process on each available CPU core, with PM2 automatically load-balancing requests between them. To ensure it survives reboots, you would then run pm2 startup followed by pm2 save.