direct-digitalocean.md

Direct deployment on Digital Ocean

Since version 0.9 nscale supports any cloud provider that give SSH access, plus bare-metal deployments.

In this tutorial we will use Digital Ocean. If you do not have an account, signing up through that link gives you 10$ of credit, enough to run through this tutorial.

In this tutorial we will deploy our beloved startup death clock application on a single machine, plus a nscale deployment machine.

Setting up the deployment machine

Create a new Droplet in DigitalOcean, and select the Docker application, like in the following screenshot:

And SSH into that machine.

(pick a location near you, for the sake of bandwidth and deployment speed it is important that all the machines are in the same datacenter)

For this tutorial, you can use the smallest tier, but for largest deployments you might need larger disks.

Install nscale

We suggest to install node via nvm, using:

curl https://raw.githubusercontent.com/creationix/nvm/v0.18.0/install.sh | bash

Then, logoff, login and run:

apt-get update
apt-get install build-essential
nvm install v0.10.33
nvm alias default v0.10.33
npm install npm@latest -g --unsafe-perm
npm install nscale -g --unsafe-perm

node.js and nscale are now installed, launch it via nscale ser start.

User configuration

You need to configure GIT to use nscale:

git config --global user.name "your name"
git config --global user.email "[email protected]"

System set up

nscale login
git clone https://github.com/nearform/sudc-system.git
cd sudc-system
nscale sys link .
nscale sys list

We need to generate an ssh key with no passphrase for this project:

ssh-keygen -t rsa

Hit enter to leave a blank passphrase, and save it as sudc-key, not as ~/.ssh/sudc-key.

Creating a new machine+

Open your Digital Ocean admin panel.

Create a new machine like we the previous one, but configure it with the newly created ssh key:

Note the IP address of the machine, for this tutorial it's: 178.62.80.16. Replace this with your own.

Test that you can access the machine from the nscale machine with:

ssh -i sudc-key 178.62.80.16

nscale configuration

In order to configure the nscale analyzer to accurately detect the containers running on the target machine, you need to edit the ~/.nscale/config/config.json file and replace:

...
"analysis": {
      "require": "nscale-local-analyzer",
      "specific": {
      }
    }
...

to:

...
  "analysis": {
      "require": "nscale-direct-analyzer",
      "specific": {
        "user": "root",
        "identityFile": "/root/sudc-key"
      }
  ...

(Adjust this config if you cloned sudc-system to another folder)

This step will likely not be needed in a next release, more goodness is coming!

Edit the system definition

Create a new file named definitions/machines.js in the project with content:

exports.machine = {
  type: 'blank-container'
};

Then, edit your system.js to the following:

exports.name = 'sudc-system';
exports.namespace = 'sudc';
exports.id = '62999e58-66a0-4e50-a870-f2673acf6c79';

exports.topology = {
  local: {
    root: ['doc', 'hist', 'real', 'web']
  },
  direct: {
    machine$123: {
      contains: ['doc', 'hist', 'real', 'web'],
      specific: {
        user: 'root',
        identityFile: '/root/sudc-system/sudc-key',
        ipAddress: '178.62.80.16'
      }
    }
  }
};

(you can specify more than one instance on the same host postponing $<identifier>)

Compile your topology with:

nscale sys comp direct

Deploying!

You can build all your container with:

nscale cont buildall sudc-system latest direct

Go grab a cup of coffee, while nscale builds everything for you.

Then, launch:

nscale rev dep sudc-system latest direct

To deploy the latest revision.

Enjoy sudc at http://YOURIP:8000/.

You can see the result of this tutorial at: https://github.com/nearform/sudc-system/tree/direct