advanced_virtual_machines

Advanced Virtual Machines

Table of Contents

• Basic Clustering
• Building the Instance Image
• Building the Instance Template
• Building the Cluster
• Balancing the Cluster

Basic Clustering

This is something we will explore under more serverless-style cloud computing Modules, but to keep it simple, clustering is important. What happens if you VM goes down? Your service goes down. This can't happen. The show must go on! In order to reduce the chance that your entire service goes down, we use clustering. Clustering is essentially a bunch of machines, called workers or nodes or slaves that are hidden behind a load balancer. The client connecting to the cluster then hit the dedicated IP or domain of the load balancer, which determines which worker to send the connection to. Then the worker sends the response back to the load balancer, which returns it to you. Let's look at a diagram:

As you have have guessed from the above diagram, we are going to be building this cluster on GCP. It is easier for GCP, and we'll be making a cluster that is fully production grade.

Building the Instance Image

In order to make the cluster, we have to start by making a disk image with our code already on it and running, that we can copy to other machines. So lets go ahead and make a near default Compute Engine instance with only these changes:

• Disk image: Ubuntu 18.04
• Allow HTTP Traffic
• Allow HTTPS Traffic

Once that is created, ssh in.

• Make a folder, and run npm init -y in it to initialize the project
• Run npm i -s express to install express
• Create a file called index.js and write the following into it:

const express = require('express')
const os = require('os')

const app = express()

app.get('/', (req, res) => {
    res.send(os.hostname())
})

app.listen(80, () => {
    console.log("Running on port 80")
})

Now we need to get that code onto the VM, in order to do this we need some carefully crafted commands:

• scp index.js root@[your ip]:~/
• scp package.json root@[your ip]:~/

Now ssh back into the VM, we gotta run some more commands:

• curl -sL https://deb.nodesource.com/setup_10.x | sudo -E bash -
• sudo apt-get install -y nodejs Those two will install Node.js 10.x
• npm i to install dependencies
• sudo node index.js to run the app (we need sudo to run on any port below 1024, we are using 80)

Now, in your browser, navigate to http://[your ip] and you should see the name of your instance

We need to install pm2 now, which will handle making sure our app runs on boot:

• Run sudo su so we can run as root
• Run npm i -g pm2
• Run pm2 start index.js --name testApp This will have pm2 manage our app
• Run pm2 ls to see the running process
• Run pm2 startup this will enable pm2 to run on boot
• Run pm2 save this will save the current processes to be ran when pm2 starts

Now you can exit from the SSH session and head over to the Compute Engine panel in GCP.

We want to go ahead and shut down the VM so we can save the instance image. Click the check box next to the image, then stop in the top menu.

Once the VM stops, click on Images in the left menu (or Compute Engine > Images)

To create an image from our instance, click CREATE IMAGE in the top menu. Give it a name, and for the source disk select your instance. Then click create. This will take a bit.

Building the Instance Template

Now we need to make the template that our cluster workers will be made from. Go to Compute Engine > Instance templates and click Create instance template

The only thing to change is the disk: Make it Custom Images > [your image name]

Don't forget to allow HTTP and HTTPs traffic as well, and click Create

Building the Cluster

Go to your GCP project and navigate to Compute Engine > Instance groups, and click Create instance group

Select the instance template, and set the maximum number on instances to 3, and minimum instance to 2 (to save some money). Make sure Autoscaling is on.

Now click Health check > Create a Health Check and set the following:

• Protocol: HTTP
• Port: 80
• Request path: /

Click save and continue

now click create

Balancing the Cluster

Now we need to create the load balancer, which will also give us HTTPS (really important!). To do so navigate to Network Services > Load balancing and click Create load balancer

Click Start configuration under HTTPS load balancing and then Continue in the first question. Do the following:

• Backend configuration
  • Backend Services
    • Create a backend service
      • in the instance group dropdown select the instance group you made, select the health check you made and click create
  • Frontend Configuration
    • protocol: http
    • port: 80
    • Click Done

Click Create

Now click on your load balancer, and copy in the IP. After waiting for like 10 minutes, navigate to http://[yourip] in the browser, and you should get a response.

Now keep refreshing, it should change as you get routed to the different workers in the cluster.

BOOM! You just made a production ready VM cluster pretty dang fast. We didn't do HTTPS because I didn't want you to have to wait for the certificate to validate before we could use the cluster, but that shouldn't be something I need to walk you through since in the menu of the load balancer creation it does a good job of that.

Don't forget to delete everything you made, the image, the instance template, instance group, instances, static IPs, and load balancer backend (different tab in load balancer menu), load balancer so you don't get caught in billing