This post is WIP.

Introduction

K8S has become more than a way to run pods. It had become a ubiquitous API layer for cloud resource management.

Mature, well tested controllers provide life cycle hooks for all CRUD operations of a resource (or multiple resources in some cases)

Let’s keep in mind, K8S is a REST api on steroids. It allows extending the core API with CRDs and custom controllers by enforcing a common pattern on all APIs that are deployed within it.

If you have worked with any cloud provider, you have always interacted with a REST API.

My proficiency is with AWS, but the concept still stands for Other Cloud Providers.

Through the UI (console) the web browser generates API requests that turns into Cloud Resources. Through the CLI, same thing with less RAM and less JavaScript (most of the times).

A common pattern emerged is to use a State Management tool to simplify our REST API calls, dependency management, and most importantly, state of our resources.

This movement is now called GitOps, but it had many names & shapes before.

(Put Example Here)

Cloud Native Management With all that theory out of the way, let’s talk practice.

Let’s first see how the tools we are discussing are similar

Similarities:

• Package management / Versioning - code reusability across projects
• Templating - Variable substitutions in a deployment
• CREATE, READ, UPDATE REST API calls (we will discuss DELETE later) - Operations against an API

Differences:

• State Management
• Dependency resolution

Helm

The Helm project has been used from almost the beginning of the K8S project.

The current recommended (and secure) version is Helm 3.

Helm uses Golang’s templating features to provide variable substitution.

State management is achieved by Helm adding metadata tags on each API resource it creates, and queries the K8S API to maintain state.

Helm is very popular and is considered a standard tool in many Cloud Native scenarios.

Kustomize

If you want to join YAML files together for multiple environments, and like using straight up kubectl, look no further than Kustomize.

Kustomize’s power comes from its simplicity. It doesn’t manage state, doesn’t handle type safety, or even powerful templating.

It is purely a data structure merger. Want to merge 2 objects? Merge lists? Override lists? Merge nested data structures? Kustomize probably has a solution for you.

Terraform

At a high level, Terraform is a templating language and state management solution for the Cloud. It’s power comes from its providers.

Unlike the tools listed here, Terraform was created to make manage the dependencies of Cloud Resources.

This means automatically resolving which resources depends on the creation of other resources.

One thing to note when using Terraform is that you can use multiple providers to support a vast amount of use-cases.

Tanka

From Grafana we get this cool project.

Instead of YAML/JSON manipulation, Tanka takes the approach of using a templating language - Jsonnet.

Jsonnet unlike YAML is a programming language with a runtime.

That means we can leverage our development experience from other languages, and apply it to a new language.

Similar to Helm, Tanka