Skip to content

Spinnaker Pipelines (and other) as Code using github.com/karlskewes/spin-libsonnet

License

Notifications You must be signed in to change notification settings

karlskewes/spinnaker-pac

Repository files navigation

Spinnaker Pipelines as Code

An opinionated framework for managing Spinnaker Applications, Pipelines and Projects as Code using Jsonnet and the spin-libsonnet library.

Target audience

Designed for Platform teams operating Spinnaker to codify infrastructure concerns whilst providing an application and pipeline DSL for Product teams to self-serve.

This project expects the Platform team(s) to own a monorepo containing all jsonnet and review pull requests made by Product teams for their applications. Jsonnet enables off-roading through composition which is both powerful and potentially undesirable. For example, targeting a different AWS account or Kubernetes cluster.

Goals

  • Language - user focused DSL - artifacts, environments, ordering
  • Consistency - deploy any application with the same pipeline building blocks
  • Programmatic - no manual clicking through UI
  • Lightweight - easy for user to onboard new app with minimal jsonnet

The deploy pipeline

The deploy.jsonnet pipeline is the special sauce and also where the complexity lies.

It provides a golden path that is easy to onboard to whilst supporting ad-hoc pipeline extensions to cope with the messy reality of codifying business processes.

The example structure generates a single Spinnaker pipeline per Spinnaker application.

I believe that a single pipeline deploying through any and all environments to production is preferable.

A single pipeline is simpler to operate, results in less drift across environments if more than one, becomes familiar across teams and departments, accomplishes shorter time-to-value and lower failure rates.

A single pipeline does put tension on processes because new artifact versions can pile up waiting for the existing pipeline to complete.

That all said, you can add your own pipeline templates and wire up applications to use them.

Getting started

Whether you're in a Platform team operating Spinnaker or in a Product team using Spinnaker as a customer, start with the TUTORIAL.md.

This will give you a basic understanding of working with jsonnet and the key features you might need.

Example implementation

After exploring the tutorial have a browse through the example/ directory.

cd example/

Render all jsonnet to json:

make build

Deploy all applications, pipelines and projects:

make deploy-all

Other make targets:

$ make
build                          Generate JSON manifests for `spin save`
clean                          Removes any manifests from previous builds
dep                            Install dependencies
deploy-all                     Deploy all applications, pipelines, projects
deploy-applications            Deploy application manifest(s) to Spinnaker - default all, else use 'FILE=path/to/file.json'
deploy-pipelines               Deploy pipeline manifest(s) to Spinnaker - default all, else use 'FILE=path/to/file.json'
deploy-projects                Deploy project manifest(s) to Spinnaker - default all, else use 'FILE=path/to/file.json'
test                           Test for rendering, formatting and linting errors

Contributing

See CONTRIBUTING.md

About

Spinnaker Pipelines (and other) as Code using github.com/karlskewes/spin-libsonnet

Resources

License

Stars

Watchers

Forks

Packages

No packages published