Skip to main content
Table of contents

Infrastructure

Provision machines for data science research

Data science research frequently requires high-powered machines for limited periods of time.

Machines for this purpose are created in the govuk-tools AWS account using the app-data-science Terraform project.

Note

The machines provisioned using this method are usually high powered and costly. Therefore, please ensure they are destroyed once the research has been completed.

 Autoscaling

Machines created using this project are created in autoscaling groups that ensure at least 1 instance of each machine is running at all times.

Once research is finished, machines should be destroyed by removing them from the project and re-deploying it. Alternatively, if the machine will be required again in a short period of time, the autoscaling group can be manually set to zero instances in the meantime.

SSH access

All machines have public IP addresses and can be accessed via SSH. The userdata for each machine is set up to add SSH keys for the people who will be accessing them.

This userdata should be updated when people need to be added or removed. Existing machines will then need to be re-deployed (most quickly by destroying the existing instances and allowing the autoscaling group to re-create them).

SSH access to data science machines is not controlled by Puppet or AWS access.

This page was last reviewed on 5 September 2019. It needs to be reviewed again on 5 March 2020 by the page owner #govuk-2ndline .
This page was set to be reviewed before 5 March 2020 by the page owner #govuk-2ndline. This might mean the content is out of date.