Skip to content

Latest commit

 

History

History
309 lines (235 loc) · 11 KB

README.md

File metadata and controls

309 lines (235 loc) · 11 KB
Raw

kustomization-base

This repo contains a base kustomization that can be used to install Spinnaker to a Kubernetes cluster. Please see the RFC for an overview of the new Kubernetes-native installation pathway that is under development.

This repo is under active development and is not ready for production (or even development) use, but stay tuned for updates!

Installing Spinnaker

Prerequisites

Before you start, you will need:

  • A Kubernetes cluster and kubectl configured to communicate with that cluster
  • The kustomize binary installed on your machine. Note that the version of kustomize bundled with kubectl is out of date and will not work with these kustomizations.
  • The kleat binary installed on your machine

Instructions

Fork the spinnaker-config repository, and clone this fork to your local machine. This repository contains a template kustomization that you will configure to deploy Spinnaker.

Change to the base directory; all further commands will assume that you are in this directory.

At this point, you can generate the YAML needed to deploy Spinnaker by running kustomize build .. As we have not configured any of the services, many would fail on startup (particularly those that depend on some configured persistence store), so let's add some configuration.

Create a spinnaker namespace

Create a new namespace called spinnaker into which each microservice will be deployed:

kubectl create namespace spinnaker

If you would like to use a different namespace, you will need to update the namespace in your root kustomization.yml and each microservice's baseUrl in a fork of this repository's service-discovery/spinnaker.yml.

Configure Redis

A number of Spinnaker services require a Redis connection; this install pathway requires there to be a service redis in the spinnaker namespace.

In the most common case, you will have an external redis; the template repo has a Service configured with an ExternalName; you can update this to point to the DNS name of your redis service.

For more complex cases, please refer to the following blog post on best practices for mapping external services. In general, the only requirement of your solution is that you have a service named redis in the spinnaker namespace that routes to a valid redis backend.

Regardless of the approach you choose, add all the relevant redis Kubernetes objects to your customization via kustomize edit add resource redis/*.yml.

Add any secret files

The secrets folder is intended to hold any secret files that are referenced by your config files. The files in this folder will be mounted in /var/secrets in each pod.

To add a secret, put it in the secrets/ folder and add it to the files field of the spinnaker-secrets entry in the kustomization.yaml. In this example, we'll add a kubeconfig file called k8s-kubeconfig:

cp <path to kubeconfig> secrets/k8s-kubeconfig

and update the secret in the kustomization.yaml file as:

- behavior: merge
  name: spinnaker-secrets
  files:
    - secrets/k8s-kubeconfig

Generate Spinnaker config files

This installation pathway expects the config files for your services to be in the kleat directory. As the name suggests, the expectation is that these files will be generated by kleat and written to this directory.

kleat takes as input a halconfig. This file is in general compatible with the file consumed by Halyard; if you are migrating from Halyard, please read the minor changes you will need to make to your file in the kleat readme.

Unlike Halyard, kleat allows you to specify only the fields that you'd like to configure. In this case, we will assume a very simple config:

providers:
  kubernetes:
    enabled: true
    accounts:
      - name: k8s
        kubeconfigFile: /var/secrets/k8s-kubeconfig
    primaryAccount: k8s

Note that the kubeconfigFile argument here points to a file relative to /var/secrets. This is important, as the kleat config requires that all file references be relative to where they will be mounted in the running container (rather than where they happen to be on the machine running kleat).

Save this file as halconfig.yml to the current directory, Invoke kleat generate to consume the input halconfig.yml and output the service configs to the kleat/ directory:

kleat generate halconfig.yml kleat/

This will generate service configs in the kleat/ directory, replacing the empty configs that were there before. The kustomization.yaml file is already configured to mount these configs in the correct location in each microservice. For example, it has an entry for the clouddriver config as:

- behavior: merge
  files:
    - kleat/clouddriver.yml
  name: clouddriver-config

Enable optional services

The optional Spinnaker services this workflow currently supports are Fiat, Kayenta and Keel.

In the following kustomization files, we recommend including ?ref=${KUSTOMIZATION_BASE_REF} to pin kustomization-base to a specific version. For further details, see (Optional) Use a specific version of kustomization.

To enable Kayenta, ensure that canary.enabled is set to true in your hal config, and then ensure the Kayenta kustomization is included in the resources block of your kustomization.yml:

resources:
- github.com/spinnaker/kustomization-base/core?ref=${KUSTOMIZATION_BASE_REF}
- github.com/spinnaker/kustomization-base/kayenta?ref=${KUSTOMIZATION_BASE_REF}

To enable Fiat, ensure that security.authz is set to true in your hal config, and then ensure the Fiat kustomization is included in the resources block of your kustomization.yml:

resources:
- github.com/spinnaker/kustomization-base/core?ref=${KUSTOMIZATION_BASE_REF}
- github.com/spinnaker/kustomization-base/fiat?ref=${KUSTOMIZATION_BASE_REF}

Since the default service discovery file defaults Fiat to disabled, you will need to configure a service discovery overrides file in your fork of spinnaker-config with the following block:

services:
  fiat:
    enabled: true

To enable Keel, ensure that managedDelivery.enabled is set to true in your hal config, and then ensure the Keel kustomization is included in the resources block of your kustomization.yml:

resources:
- github.com/spinnaker/kustomization-base/core
- github.com/spinnaker/kustomization-base/keel

Set the Spinnaker version

By default, this base kustomization deploys the master-latest-validated version of each microservice, which is most recent version that has passed our integration tests. To deploy a different version, you'll need to override the version of each microservice in the images block in the kustomization.yml file.

To deploy a specific version of Spinnaker, override each image's tag with spinnaker-{version-number}. For example, to deploy Spinnaker 1.21.0, override the tag for each microservice to be spinnaker-1.21.0:

images:
  - name: us-docker.pkg.dev/spinnaker-community/docker/clouddriver
    newTag: spinnaker-1.21.0
  - name: us-docker.pkg.dev/spinnaker-community/docker/deck
    newTag: spinnaker-1.21.0
# ...

This workflow has only been tested with Spinnaker >= 1.21.0. Deploying an earlier version of Spinnaker using this workflow may not work as expected, since you may have been relying on Halyard to supply configuration defaults that were only added as microservice-level defaults after the 1.20 release branches were cut.

Replace Gate's readiness probe

If you have not enabled SSL for Gate, override Gate's readiness probe with the following patch:

readinessProbe:
  $patch: replace
  exec:
    command:
    - wget
    - --no-check-certificate
    - --spider
    - -q
    - http://localhost:8084/health

Reference the patch in your base kustomization.yml by adding the following to a patches block:

- target:
    kind: Deployment
    name: gate
  path: path/to/my/readiness/probe/patch.yml

(Optional) Use a specific version of kustomization

With a reference of the version, you can use a specific version with conviction, after examining if the version works well with your configurations. Without a reference, a resource link always references master. You can check out the available versions here.

For example:

resources:
- github.com/spinnaker/kustomization-base/core?ref=v0.1.0

For further details, see the documentation for the resources field.

(Optional) Add any -local configs

In addition to the main service.yml config file, each microservice also reads in the contents of service-local.yml to support settings that are not configurable in the halconfig.

If you would like to add a -local.yml config file for any service, add it to the local/ directory, and update that service's config in the kustomization.yaml to also mount that local.yml file.

For example, to configure for clouddriver, add these settings to local/clouddriver-local.yml, and update the clouddriver-config entry in the kustomization.yaml to:

- behavior: merge
  files:
    - kleat/clouddriver.yml
    - local/clouddriver-local.yml
  name: clouddriver-config

(Optional) Enable monitoring

The Spinnaker monitoring daemon runs as a sidecar in each Deployment (excluding Deck). To enable monitoring, copy the monitoring directory from this repository into the base directory of your fork of spinnaker-config.

Add the monitoring directory to your base kustomization.yml's resource block. This will pull in the kustomization.yml that includes configuration that each microservice's monitoring sidecar will use to discover the endpoint to poll for metrics.

Next, copy the example patches block into your base kustomization.yml. These patches will add the monitoring sidecar and appropriate volumes to each Deployment.

To include custom metric filters, add them to the included metric-filters directory in your fork of spinaker-config, and reference them in the spinnaker-monitoring-filters secretGenerator entry in the root kustomization.yml.

Deploy Spinnaker

Now that all of the config files are in place, you can generate the YAML files to install Spinnaker by running kustomize build .. You can either save this to a file and apply it, or directly pipe it to kubectl via:

kustomize build . | kubectl apply -f -