Overview

HashiPOC stands for Hashicorp Proof Of Concept. This example uses Consul, consul-template, Vault, and Nomad to deploy a sample python application, load balanced with haproxy.

Prerequisites

• A working Vagrant installation

Quickstart

Bring the vagrant instance up.

vagrant up

At the end of the provisioning two instances of the python application will be running via nomad behind haproxy. Since port 80 is mapped locally at port 8080, you can curl the application locally:

while [ 1 ] ; do curl http://localhost:8080 ; sleep 1 ; done

Alternatively, you can SSH into the vagrant instance and run the while loop there:

vagrant ssh -c 'while [ 1 ] ; do curl http://localhost ; sleep 1 ; done'

Also installed is a UI for viewing and manipulating consul and nomad. Point your browser at http://localhost:3000

Under the hood

After bringing the Ubuntu Xenial instance up, shell-based provisioning happens.

base.sh installs some basic tools: unzip, supervisor, virtualenv, pip, and installs a pip requirements file.

haproxy.sh installs haproxy.

consul.sh, haproxy.sh, vault.sh, nomad.sh, and hashi-ui.sh install those five packages with supervisor config files to manage their processes. Some vault token generation, policy generation, and secret creation also happens in vault.sh.

consul-template.sh installs consul-template, which monitors service changes in consul and reconfigures and reloads haproxy accordingly.

dnsmasq.sh exposes consul's DNS so that you can do dig helloworld.service.consul +short SRV instead of having to point at consul's DNS to do service discovery.

application.sh initiates the nomad command to run the Hello World python code.

Doing a rolling upgrade

The nomad deployment (files/helloworld.nomad locally, /vagrant/helloworld.nomad on the vagrant instance) is configured to do a blue/green style release. Inside this file, change "v1" to "v2" on line 19. Keep a while loop curling localhost going in another window. Then execute nomad run /vagrant/helloworld.nomad on the vagrant instance. You'll see four instances of the application running instead of two (two of which will print out v2). If you execute a nomad deployment list and grab the deployment id, you can promote it with nomad deployment promote <ID>. The old instances will go away, leaving the v2 instances in place.