# Infrastructure Documentation

This folder contains our infrastructure documentation. These documents may be
of interest for the team that maintain the Tekton own CI/CD setup as well as
for anyone interested in using Tekton to run (part of) their own CI/CD
infrastructure.

- [Clusters](#clusters)
- [GCP Projects](#gcp-projects)
- [DNS](#dns)

## Clusters

The infra system relies on several different kubernetes clusters, three are
static and the rest are dynamic (provisioned on demand).

- [*prow*](prow.md): Prow, Boskos and Tekton run in this cluster.
  This cluster runs resources defined in the `prow` folder. CI Jobs that only
  require a container run in the `test-pods` namespace of this cluster.
- [*dogfooding*](dogfooding.md): Tekton runs in this cluster. This cluster is
  setup with [resources](../tekton/README.md#resources-for-cicd) from the
  `tekton` folder, plus a few [secrets](./dogfooding.md#secrets).
- [*robocat*](robocat.md): This cluster is our test bed for continuous
  deployment of services and resources. Everything that runs in this cluster is
  deployed automatically, which means it must be possible at any time to delete
  the cluster and recreate it from scratch.

## GCP projects

Automation for the `tektoncd` org runs in a GKE cluster which
[members of the governing board](https://github.com/tektoncd/community/blob/main/governance.md#permissions-and-access)
have access to.

There are several GCP projects used by Tekton:

- The GCP project that is used for GKE, storage, etc. is called
  [`tekton-releases`](http://console.cloud.google.com/home/dashboard?project=tekton-releases). It has several GKE clusters:
  - The GKE cluster that is used for [`Prow`](prow/README.md), `Tekton`, and [`boskos`](boskos/README.md) is called
    [`prow`](https://console.cloud.google.com/kubernetes/clusters/details/us-central1-a/prow?project=tekton-releases)
  - The GKE cluster that is used for nightly releases and other dogfooding is called
    [`dogfooding`](https://console.cloud.google.com/kubernetes/clusters/details/us-central1-a/dogfooding?project=tekton-releases)
- The GCP project
  [`tekton-nightly`](http://console.cloud.google.com/home/dashboard?project=tekton-nightly)
  is used to hold nightly release artifacts and [the robocat cluster](#the-robocat-cluster)

### Adjusting GCP permissions
Access and permissions to the GCP projects are automated via [Terraform](https://www.terraform.io/). The configuration is maintained in the [tektoncd/infra](tektoncd/infra) repo. The [@plumbing.collaborators](https://github.com/orgs/tektoncd/teams/plumbing-collaborators) team has read access and [@plumbing.maintainers](https://github.com/orgs/tektoncd/teams/plumbing-maintainers) team has maintainer access to this repository. To adjust permissions, please ask in the #plumbing channel on [Slack](https://github.com/tektoncd/community/blob/main/contact.md#slack). 

## DNS

DNS Names are managed via [Netlify](https://www.netlify.com/).
[Gardener External DNS Manager](https://github.com/gardener/external-dns-management) is deployed on the [dogfooding](./dogfooding.md#dns-names) and robocat clusters, and manages names
via annotations on ingresses and services. Some of the names are defined manually in Netlify.

## End-to-end testing

We are currently shifting from using GKE clusters for end-to-end testing of PRs to using 
`kind`. See [this document](./kind-e2e.md) for information on moving your existing 
end-to-end CI jobs to use `kind`.