run components of a large project on separate machines
Place your existing node application in a sub directory, and create a
server.js file so it is relative to the other files like this. Let's assume we started with two node modules or servers that work together (
application) We handled deployment and version control of these applications separately because, well, we had to, but it's about to get a lot easier.
├─┬ api│ ├── server.js│ └── .git├─┬ application│ ├── server.js│ └── .git├── .git├── deploy.js├── network.json└── server.js
Create a very simple network.json file. The important part when creating components here is to name the directory that holds the component the same as in the
Now add to the new server.js
var host_component =var network =
Before you deploy, set your hostname so it contains the name of the component you want to run on the machine. Avoid overly generalized names and collisions.
Since our hostname contains
api and this matches with our
network.json and subdirectory named
api -- a call to
node server will trigger the
api component to load. Now you can deploy the same source code to every machine for maximum DRYness.
Let's setup the application machine too.
Ok great, now lets go back and update our network.json file so it looks something like this. Now you can feel free to add more components to your system without a lot of complexity. Just remember to choose hostnames wisely.
"api":"spinoza":"public": "spinoza-api.joyent.us""aquinas":"public": "aquinas-api.joyent.us""socrates":"public": "socrates-api.joyent.us""application":"locke":"public": "locke-application.joyent.us"
Finally, we deploy to all of our machines with a simple
node deploy command using another npm module
var deploy =