GridControl

GridControl provision and links multiple servers togethers to form a Grid.

Files are synchronized, Opionated Pub/Sub system is set, Servers get linked together.

You develop, you play, in a scalable way. The more Servers you add to the Grid, the more calculation power you get.

Welcome to the Grid

5 minutes to get started. By the authors of PM2.

Behind the scene: GridControl is a network layer built on top of PM2 allowing file synchronization, inter process communication via an opionated PUB/SUB system and a wide-range system discovery

Features

• 0 conf Auto discovery system via multiple DNS
• 0 conf P2P application source sharing
• Ecosystem Grid management toolbox (CLI, provisioning, Logs, Monitoring)
• Secure Diffie Hellman key exchange and password authentication
• Decentralized Each Nodes can trigger actions executed by another Nodes
• Fast Grid interconnected via TCP sockets
• Fast Services are started once, then stay alive waiting for inputs. This saves non-negligible startup time
• Polyglot Services can be wrotte in any language
• Compatible with Amazon Lambda, Google Cloud Functions
• Rock Solid PM2 behind the scene for process management and cluster capabilities
• And a lot more like Buffering, Retry on Failure...

Creating a Grid

Install your Army Swiss Knife to manage a Grid:

$ npm install grid-cli -g

Now the bin grid is available via the CLI.

Commands

1/ Generate a new Gridfile in the current path that contains grid name, grid password, host and SSH keys:

$ grid new

The Gridfile will look like this:

grid_name     = 'grid-name'
grid_password = 'xxxx'

servers = [
  'user@ip1',
  'user@ip2',
  'user@ip3'
]

ssh_key = '''
1024_PRIVATE_SSH_KEY
'''

ssh_public_key = '''
PUBLIC_SSH_KEY
'''

Change each attribute with the desired value.

2/ Provision every hosts listed in the Gridfile:

$ grid provision

This will copy the SSH pub key and install NVM, Node.js, PM2 and Gridcontrol This installation does not need ROOT access rights at anytime

3/ Grid management

# List all nodes linked to the grid
$ grid ls

# Display Dashboard
$ grid dash

# Execute a command on each server
$ grid multissh <bash_command>

# Restart/Recover the current Grid
$ grid restart

# Upgrade Gridcontrol to latest version
$ grid upgrade

# Display realtime logs of all tasks
$ grid logs

# Monitor the whole Grid with Keymetrics
$ grid monitor <secret> <public>

# Interactively SSH to desired machine
$ grid ssh

Interact with the Grid

Now let's play with the Grid. You can generate a sample project by typing:

$ grid sample [project-name]
$ cd [project-name]
$ npm install

You will get a project looking like this:

.
├── index.js
├── package.json
└── tasks
    └── get-ip.js

Let's look at the content of tasks/get-ip.js, this is a task that will be propagated in the grid:

var request = require('request');

module.exports = function(context, cb) {
  request('https://api.ipify.org/?format=json', function (error, response, body) {
    if (error)
      return cb(error);

    if (!error && response.statusCode == 200) {
      return cb(null, body);
    }
  });
};

To call this function, look at how it is done in index.js:

var grid = require('grid-api').init({
  instances   : 1,
  env         : {
    NODE_ENV  : 'development'
  }
});

function triggerTask() {
  grid.exec('get-ip', function(err, data, server) {
    if (err) {
      console.log(err);
      return false;
    }
	  console.log('Got response from server pub_ip=(%s) priv_ip=(%s):',
                server.public_ip,
                server.private_ip);
    console.log(data);
  });
}

grid.on('ready', function() {
  console.log('Gridcontrol Ready');
  setInterval(triggerTask, 1000);
});

Start the main application:

$ node index.js

At the beginning, only the local gridcontrol will respond. Once the other peers are synchronized they will also process the queries:

From server alor-vital:88.123.12.21
Got response $HTML
From server xtreme-ventage:5.10.123.144
Got response $HTML
From server veolia-graphia:88.125.120.20
Got response $HTML
From server hector-castor:120.12.1.145
Got response $HTML

Distributed processing, on-premise!

Advanced documentation

Ignoring files for synchronization

Avoid re synchronizing the application in case of some file changes by adding a .gridignore file, containing a list of regex/files to ignore separated by a newline.

Disable local tasks computation

To make the head node act as a load balancer only pass the local : false in grid.init options.

Contributing

Please refer to doc/CONTRIBUTING.md

License

Apache V2 (see LICENSE.txt)