MCP DriveTrain and upgrades: a few questions and answers

Mirantis Field CTO Ryan Day recently sat down to demonstrate how users can use Mirantis DriveTrain to upgrade OpenStack over two versions, from Mitaka to Ocata.  Along the way, he answered a number of different questions about how it works, and what DriveTrain is all about.  You can view the entire demonstration, and we've decided to provide the answers here as well.

Q: Can I use salt to deploy from bare metal?
A: You can use MaaS, Ironic, Foreman, and so on to install an operating system on bare-metal, as well as to install a salt minion agent. Once you get to that point, salt takes over configuration. Also, currently in tech-preview is salt automation of MaaS bare-metal provisioning.
Q: Are OpenStack and Kubernetes are deployed in containers?
A: Some DriveTrain components are deployed in containers. Kubernetes and OpenStack are installed with packages today, but containerized OpenStack services are in Mirantis' product roadmap.
Q: Does DriveTrain follow devops best practice pipelines using open source?
A: I'm assuming this question is about Mirantis remaining an open-source company. The answer is yes, we package and harden open-source cloud platforms such as OpenStack and Kubernetes, and also leverage open-source components for DriveTrain and Stacklight. Salt formulas are also open sourced at http://github.com/salt-formulas.
Q: Does DriveTrain also do LCM for the new Mirantis Containers as a Service?
A: Yes.
Q: So do you need replacement nodes for the control plane to upgrade, or is this a forklift upgrade?
A: No, but there are situations where this would be useful, such as if you want to minimize control-plane outage by running dual control-planes in parallel until the database cutover, but lack the resources on nodes hosting the existing control-plane.
Q: How about upgrading from old versions of Mirantis OpenStack such as Grizzly or Icehouse) to the latest version?
A: While it is theoretically possible depending on our support for a given service formula (i.e. nova, glance), we have only only tested, and claim support for, Mitaka (MCP 1.0) and newer.
Q: Why Salt and not Ansible instead?
A: Salt has certain benefits around scale, speed and architecture. You might want to look online for some comparisons.
Q: Is DriveTrain a product Mirantis sells? What does it take to re-use it in another environment, only metadata changes?
A: DriveTrain is one component of MCP as a product. Adapting DriveTrain to orchestrate another applications or services is neither trivial nor difficult, and in fact DriveTrain is well suited to providing LCM for your internal applications. It requires a solid understanding of how salt, reclass and jenkins work.
Q: Does this pipeline include any automated tests? Specifically, are there Tempest API test cases, scenarios, or any tenant-level automated testing?
A: The pipeline shown in the demo only had basic verification tests for database schema and openstack api changes. Tempest, Rally, and performance testing has been automated but were not shown in this demo due to time limitations.
Q: Can DriveTrain be used to upgrade and manage NFV workloads on Openstack?
A: Technically yes, though in the context of NFV there will likely need to be integration with specific NVFo or VNFm (orchestrator and VNF manager respectively). We have done and continue to improve this exact type of integration with Cloudify as an NFVo.
 

Again, if you'd like to view the entire webinar, you can do that here.  Also, if you're interested in NFV or operations, please check out our upcoming ONAP webinar, or our bootcamp on top sysadmin tasks and how to achieve them in OpenStack.