Following on from vCluster update number 2 I have plenty more to share with you as we approach the private beta stage of vCluster development.
An important part of the vCluster stack is the ability to deploy the entire virtual environment very quickly and with as little manual involvement as possible. This is not only part of our long term plan to offer availability zones, but also makes us able to rapidly repsond as resource and availability requirements change.
In order to do this we have been using the Ansible deployment system. Ansible is very simple when compared to Puppet or Chef and allows all configuration over SSH so a dedicated management node is not required. Ansilbe allows us to use ‘playbooks’ to configure multiple nodes at the same time and to track changes. This is perfect for our clustered architecture. We can deploy multiple web nodes each with different variables (IP, hostname etc) from a single Ansible playbook.
Our dev-ops and engineering team have been working heavily on completing these scripts over the last few weeks and last week they completed the final few test deployments.
The production hardware environment has been built and tested and we have successfully completed multiple deployments using Ansible. We are currently working on a couple of minor issues around networking but we are confident they will be resolved in the next couple of days.
Our long term plans, as I mentioned before, include multiple availability zones and we will look at building and deploying vClusters in the various data centres we have a point of presence in, however the current production stack is hosted at the MCH data suite.
Migrating to vCluster 2.0
We are aiming to have a stable production ready environment in the next copule of days. Following this we will migrate some of our own websites onto the new vCluster to build up billing and usage data and to test the software on the platform. To be clear we will move our sites only when we are happy the underlying stack is stable.
During this period I will offer some test accounts to our partners to gain a final round of feedback before we open for public beta. Our goal is to move to public beta as rapidly as possible following the private beta month, but only when we are 100% content that our billing data is being represented in fair and competitively priced packages, and that the vCluster has been stable and reliable during the private beta phase.
As usual I will write further updates as we continue vCluster development, hopefully just one more before we open for public accounts again!