Yesterday, we migrated Mozillians.org from Mozilla’s Santa Clara data center to ParSys AWS infrastructure. This is the result of 6 months of work from everyone on our team. It included the migration of
- the Reps portal
- a huge amount of work to modernize mozillians.org
- making changes to the infrastructure to support the security model that is required for Mozillians.org.
Below are some cool highlights from the journey.
One of the side goals of this migration was to scale up our production cluster and move our staging sites to it. Because we’re using Terraform, most of our infrastructure is defined as version-controlled code. This meant we could recreate the supporting resources for mozillians and other sites by changing a couple of lines, saving us a whole bunch of time:
A cluster of clusters
Since mozillians.org will become the backbone of the ‘A’ in the IAM project, we decided to completely separate the production mozillians.org from the other sites in our infrastructure. Thanks to the magic of Mesos attributes, we are able to attach some information about each server to use in deployment, such as region, instance size and most importantly, role. From there, we can simply tell mozillians and other apps where they’re allowed to be deployed:
We’re planning to add flower to monitor some of the background tasks that mozillians performs, along with re-configuring our general monitoring to show better and more useful information from the new cluster. We’re also going to perfect our deployment pipeline by adding better notifications and improving the scripts that interact with Marathon.