Goal

Instance types and preferences provide a way to define common VirtualMachine workload resource sizing and preferences.

The common-instancetypes project within KubeVirt currently provides a set of standardized instance types and preferences that can be deployed by virt-operator alongside all of the other core KubeVirt components.

While the resources provided by the common-instancetypes project are a great starting point they are generalized and could never possibly cover all workload resourcing and preference use cases.

As such this design proposal seeks to make it as easy as possible for both cluster admins and users to create their own customized versions of these commonly deployed resources specific to their workloads. Additionally the instance type API itself will be made more flexible by making vCPU and memory resource sizing optional within an instance type, allowing users to provide their own sizing within their VirtualMachine while still retaining the remaining benefits of using an instance type.

User Stories

• As a VM owner I want to provide my own vCPU and memory resource sizing while using commonly deployed instance types
• As a VM owner I want to create customized versions of commonly deployed instance types and preferences specific to my workload use case
• As a cluster admin I want users to be able to provide their own vCPU and memory resource sizing when using commonly deployed instance types
• As a cluster admin I want to create customized versions of commonly deployed instance types and preferences specific to my workload use case

Non-Requirements

Notes