Juju is a new way to automate cloud infrastructure using easy-to-use commands as well as a web GUI. It enables you to easily deploy your applications to a variety of cloud-like management architectures, including EC2, OpenStack, Azure, and your local Ubuntu installation. Central to Juju's distributed systems magic is the notion of a Charm---an application-specific asset that defines how to keep services running.
Using Juju, we can easily setup and administer an AWS environment for HyperDex. Juju makes this process so easy that we can literally go from nothing to a ready-to-test HyperDex cluster in just 10-minutes.
To start with, you'll need to install Juju. On Ubuntu, this is just two commands:
$ sudo add-apt-repository ppa:juju/stable
$ sudo apt-get update && sudo apt-get install juju-core
Installing on OS X is similarly easy:
$ brew install juju
Now that you've got Juju installed, let's point our Juju install at our AWS account. You'll need your AWSAccessKeyID and AWSSecretKey that you can get from your AWS management console. Run the following command to generate the config:
$ juju generate-config
$ nano ~/.juju/environments.yaml
Look for the block labelled amazon: and change the access-key: and secret-key: blocks and set them to your key id and secret key respectively. If you run into difficulty, the Juju docs can help with troubleshooting this step.
Now that we have a working Juju environment, let's use it to deploy HyperDex.
Deploying HyperDex on Juju is a three step process that fetches the HyperDex Charm, bootstraps Juju onto EC2, and finally deploys HyperDex itself. First, let's fetch the HyperDex Charm:
$ mkdir precise
$ git clone https://github.com/rescrv/juju-HyperDex.git precise/hyperdex
We now now have the local Charm ready to go with Juju. Let's bootstrap Juju onto EC2:
$ juju bootstrap
This process could take a few minutes, as it initializes a new EC2 instance to manage your Juju environment. When it completes, you can then deploy HyperDex with:
$ juju deploy --repository=`pwd` local:precise/hyperdex
Like the previous step, this will launch a new EC2 instance, and will deploy a single-node HyperDex cluster on it. This command runs in the background. To track its progress, you can use juju debug-log to see the output as the Charm installs and deploys HyperDex. When you see the following lines in the log, your cluster is ready to go:
2014-05-22 13:32:41 INFO start * Starting HyperDex coordinator hyperdex-coordinator
2014-05-22 13:32:42 INFO start * Starting HyperDex daemon hyperdex-daemon
Now that you have a HyperDex cluster ready to go, it's a great time to use it to step through the HyperDex QuickStart Tutorial. You can skip the parts of the tutorial related to setting up a cluster---it's already done for you by Juju. Login to your machine and get started with the tutorial like so:
$ juju ssh hyperdex/0
$ hostname
ip-172-31-27-22
$ python
>>> import hyperdex.admin
>>> a = hyperdex.admin.Admin('ip-172-31-27-22', 1982)
...
The only difference between this and the tutorial is that we're using the hostname of the machine as the coordinator address, not the address 127.0.0.1. Wherever you see 127.0.0.1 in the tutorial, replace it with the output of the hostname command.
Your HyperDex instance on EC2 will continue to run even after you shut down your local computer. Once you go through the tutorial, you may clean it up with one simple command:
$ juju destroy-environment amazon
WARNING! this command will destroy the "amazon" environment (type: ec2)
This includes all machines, services, data and other resources.
Continue [y/N]? y
This will shutdown HyperDex and release all associated EC2 resources.
Our Juju Charm is still quite new, and doesn't support many of the features of Juju (yet). If you would like to use this Charm in production, we recommend contacting the HyperDex team first, so that we can help make sure the Charm meets your needs. Here are some things that Juju supports that we would love to have the HyperDex Charm support as well:
Keep an eye on this blog, and on our GitHub repo for the HyperDex Charm to watch for these feature and more.