Name Company Topic: “how we at StackHPC deploy Scientific OpenStack”
Here is one way of thinking about an OpenStack cloud environment, from a very high level. We have some physical infrastructure, an OpenStack control plane running on top, some resources registered for various projects, and user workloads running within those projects. And of course, let’s not forget that all important monitoring and logging.
If you can, try to clear that image from your mind.
You’ve got mail!
Uh oh, this is quite daunting. We have a linux box, some freshly racked iron, and a spreadsheet. This doesn’t look at all like a Super Science Cloud.
Thankfully, we’ve chosen to use Kayobe to deploy our cloud. Kayobe is a tool that ties together a number of other tools to provide a complete cloud deployment solution.
Bifrost deploys a standalone ironic service, and uses it to discover and inspect the control plane hardware, then provision it with an OS.
So we install Kayobe on our machine, and use it to install Bifrost on this Linux box in the centre that we’ll call the seed. Kayobe can configure the physical network devices, if you like. Using bifrost, we can discover, inspect and provision the control plane hardware. Kayobe then configures the OS on the control plane hosts, and installs the Docker engine.
Kolla is a tool for building container images for OpenStack control plane services. Kolla ansible uses those container images to deploy an OpenStack control plane. Both tools support multiple OS distributions.
Monasca is a scalable solution for monitoring and log aggregation. It’s multi-tenant, so can monitor both the control plane and user workloads.
At this point, kayobe installs kolla on the seed, and uses it to build container images. Kolla Ansible is then installed on the operator machine, and used to deploy OpenStack and Monasca.
Ssc-config is a pattern we use whereby project resources such as images, flavors, networks, are declaratively defined using Ansible. Most of the logic is in reusable roles shared on Galaxy.
So we install ssc-config, and register the resources for various projects.
Ssc-appliances is a pattern we use for automating the creation of clusters running on OpenStack using Heat or Magnum.
So we can install scc-appliances on our machine, and start creating clusters for user workloads.
Deploying Scientific OpenStack
An OpenStack Cloud - Sky High View
OpenStack Control Plane
We’ve racked and cabled the servers and
switches for the Super Science Cloud (SSC).
The seed has been provisioned with an OS, as
I’ve attached a spreadsheet containing
Over to you!