scaleme

Movitation

Suppose you have a microservice, API or web application that you need to deploy, but you are concerned about the possibility of your server not being capable of handling it. You could use AWS to scale horizontally your servers for you, the problem (and the biggest motivation for this project) is that AWS is expensive compared to another providers, for example, Digital Ocean, Linode, OVH, etc.

Prerequisites

A Digital Ocean (for now) and CloudFlare accounts. We will use CloudFlare DNS to load balance to your machines, so, be sure that the root domain you wanted exposed in scaleme config file is signed in your CloudFlare account.

More details about the implementation can be found at the How it Works section in this page.

Installation

Node.js is necessary to run this project. You should go to the Node.js Downloads Page and proceed with their instructions.

Then, just clone this repository and start it using npm (Node package manager):

git clone https://github.com/fesnt/scaleme.git
cd scaleme/
sudo npm install -g

Getting Started

Now you can use scaleme CLI to start, list, monitor and delete your projects.

Starting a Project

To start a project, first you must create a config file in any folder you want (check out config.sm). The basic structure of this file is this:

{
	"project": "test", // project name
	"git" : "https://github.com/fesnt/extract_article.git", // docker container app
	"port" : 3000, // port that your container runs
	"digital_ocean" : {
		"key": "YOUR_DO_KEY",
		"region" : "nyc3",
		"size": "512mb"
	},
	"cloudflare" : {
		"email": "your@email.com",
		"key": "YOU_CLOUDFLARE_KEY",
		"domain": "test.example.com"
	},
	"scaler_rules": {
		"min": 1, // min number of computers
		"max": 3, // max number of computers
		"interval" : 300, // interval to monitor and decide to scale (seconds)
		"up": {
			"metric": "cpu",
			"rule": "gt",
			"percent": 70
		},
		"down": {
			"metric": "cpu",
			"rule": "lt",
			"percent": 10
		}
	}
}

Then, run:

sudo scaleme -s config.sm

Your project will start and after a couple of minutes you can check it out using the domain you provided, in this example test.example.com.

Listing current projects

sudo scaleme -l

Get monitoring information of a project

sudo scaleme -m project_name

Delete a project

sudo scaleme -d project_name

How it Works?

scaleme initially deploy two machines, one called the scaler and another the slave.

The job of the scaler is to monitor the slave(s) using the rules defined in the config file, if the up rule is triggered then the scaler will deploy another slave machine if the maximum number of machines is not reached, and if the down rule is triggered it will delete one slave machine if the mininum number of machines is not reached.

Besides, another job of the scaler is to get the slaves ips and sign them at the CloudFlare DNS so that your machines can be acessible from the domain you provided. The CloudFlare DNS act like a load balancer, since we have more than one ip from the same domain, the DNS will deliver this ips in a round-robin fashion, distributing the load among all the slaves you have at a given moment.

The job of the slave is straight-forward, it is your docker application with a monitoring API exposed to enable the scaler to request monitoring informations and make scaling decisions.

The picture below illustrates the infrastructure that will be create for you in seconds:

felipe-augusto/scaleme