Brian Grant

Fragmentation has been exacerbated by user preferences to use familiar templating and programming languages, the simplest mechanisms that satisfy their needs, and as few tools as possible. Greater reusability of tool implementations and config itself would benefit everyone.

That’s one of the many reasons I am excited about kpt. kpt is a toolkit to help manage, examine, manipulate, customize, validate, and apply Kubernetes resource configuration files, both manually and programmatically.

One of the primary sets of capabilities that is required to enable configuration reuse is the ability to distribute/publish/share, compose, and update groups of configuration files, commonly known as packages.

git is the most common source of truth for declarative configuration files, and users are increasingly driving deployment workflows from changes in git, either with traditional CI/CD tools or dedicated GitOps tools. kpt facilitates declarative workflows directly around git.

The other main capability needed to enable reuse is the ability to customize configuration files for specific use cases. That’s where the most diverse preferences are evident. kpt supports generation, transformation, and validation functions written in any programming language.

kpt accomplishes this by representing configuration as data, specifically Kubernetes resources serialized as YAML or JSON. The configuration-as-data approach builds upon the design of the Kubernetes resource model (KRM): https://github.com/kubernetes/community/blob/master/contributors/design-proposals/architecture/resource-management.md …