The requirement of this OpenStack setup is to have a highly available cloud spread over two or more data centers. It will replace the current solution where a user is requesting a VM from the infrastructure group, which is ineffective and time consuming.
The aim of the setup is to be able to lose one data center, without affecting production. The use of containers makes it easier to scale the services by adding more servers, and to set up an identical staging environment to test upgrades or develop new services.
It was also a good opportunity to introduce new technology and when you setup an OpenStack service as a container from scratch, you get pretty familiar with the service.
We are not using docker overlay network at the moment, but it would be nice to have it in the future, so we don't have to expose too many ...