The vSphere Kubernetes Drivers Operator project is designed to simplify and automate the lifecycle management of CNI, CSI and CPI drivers in a Kubernetes cluster running on vSphere. It will do this by following the Kubernetes operator model in which desired state and configuration is defined in a custom resource and the operator(s) running in Kubernetes will attempt to reconcile the drivers to satisfy the desired state.
Our goal is to make vSphere the best place to run any Kubernetes and to vastly simplify the experience of deploying Kubernetes to vSphere.
- Improve the resiliance, error reporting and state transition status of the existing drivers
- Provide a single point of reference to the information necessary to perform a manual install of the vSphere drivers
- Ensure that the existing documentation for a manual install is up to date and correct
- Ensure that common error conditions are tested, feedback from the driver is clear and remediation information is available
- Document how to contribute
- Design a consolidated configuration format - the Spec of the custom resource(s) - that covers the config needed for all of the drivers
- Come up with a design spec for the operator, including but not exclusive to:
- The topology - should it be a single operator with a controller per driver or multiple operators?
- Build depedencies - are the existing drivers statically linked into the new operator or does it load them dynamically?
- How do the drivers consume the new CRD format? Should it be cloned? Should the config format be pluggable?
- Control flow for each driver. What are the preconditions that have to be met before it can be installed?
- Interdependencies between drivers. How are state transitions managed?
- What Status would users expect to be able to see from the drivers operator?
- What CNI implementations are we supporting?
- How should we define health and liveness?
- What metrics might we want to export for monitoring?
- How is compatibility defined and enforced? Between the drivers themselves, the K8S version in use and the target vSphere
- Define common expected failure conditions and how these should be reported to the user
- Ensure that requirements for specific distributions such as OpenShift are properly captured
- Define basic driver framework, including makefiles, build and test. Include proposed framework(s) Eg. Kubebuilder
- Create controllers with stubs that assert the defined state transitions, status reporting and basic lifecycle functions work
- Build out controllers for each of the drivers incrementally ensuring integration tests are added where necessary
- End goal of Phase 2 is to have a functioning drivers operator for upstream K8S that installs and deletes drivers
- Deliver any additional requirements for specific distributions, such as OpenShift
All of the exising CPI, CSI and CNI driver projects are maintained in GitHub
We really value the community of developers and vSphere users who run Kubernetes on vSphere and our goal is to ensure that the vSphere Drivers Operator is designed and developed 100% in the open. As such, we will be using GitHub issues for tracking all of our work and GitHub markdown for our designs. We will be starting a regular call where design decisions are discussed and we commit to ensuring that decisions are well documented.
If you have an interest in contributing or submitting requirements, we'd love to hear from you!
Antrea is licensed under the Apache License, version 2.0