Consider three bare metal nodes:
- A Hadoop node with a JBOD
- A database server with RAID 10
- A High Performance Computing (HPC) compute node, with tuned BIOS parameters
Now imagine these are nova instances, deployed at different times on the same bare metal compute node, simply by using different images and/or flavors.
The OpenStack ironic team has been working tirelessly for the last few years to allow bare metal compute to be automatically provisioned, and configured to suit a variety of use cases.
Being able to configure things like BIOS and RAID via an API is a good step forward, but currently this must be done manually while a node is not in use.
In this session, we’ll describe and demonstrate Deploy Templates, the last of a series of features spanning multiple releases of nova and ironic. With Deploy Templates, we can use the traits requested by a nova flavor or image to trigger custom actions in ironic during deployment.
There are a number of different features that make this possible:
- Traits, for scheduling in nova and placement
- Ironic deployment steps
- Ironic deployment templates
And several more that make it useful:
- Ironic RAID configuration
- Ironic BIOS configuration
We'll cover the concepts underpinning this work, and how to make use of them both from the perspective of an operator and a user. Of course we'll have a demo to show it working in practice.