pctl

pctl

pctl is a cli tool for interacting with Profiles

• Usage
  • Search
  • Show
  • Install
    • Install via Catalog
    • Install via URL
    • Configuring a profile
    • Cloning repository resources
    • Architecture
  • List
  • Prepare
    • Pre-Flight check
  • Catalog service options
• Development
  • Working with profiles
    • Using local pin
    • Using dev tag pin
    • Using doki and the Makefile targets
• Release process
  • Tests
    • Configuring Integration Tests

Usage

For more information on all commands, run pctl --help or pctl <subcommand> --help.

Search

pctl can be used to search a catalog for profiles, example:

$ pctl search nginx
CATALOG/PROFILE                         VERSION DESCRIPTION
nginx-catalog-1/weaveworks-nginx        0.0.1   This installs nginx.
nginx-catalog-1/some-other-nginx        1.0.1   This installs some other nginx.

pctl search also lists all available profiles with --all option, example:

$ pctl search --all
CATALOG/PROFILE                                   VERSION DESCRIPTION
nginx-catalog-1/weaveworks-nginx                  0.0.1   This installs nginx.
nginx-catalog-1/bitnami-nginx                     1.0.1   This installs bitnami nginx.
nginx-catalog-1/some-other-nginx                  1.0.1   This installs some other nginx.
prometheus-catalog-1/weaveworks-prometheus        0.0.4   This installs prometheus.

Show

pctl can be used to get more information about a specific profile, example:

$ pctl show nginx-catalog-1/weaveworks-nginx
Catalog         nginx-catalog-1
Name            weaveworks-nginx
Version         0.0.1
Description     This installs nginx.
URL             https://github.com/weaveworks/nginx-profile
Maintainer      weaveworks (https://github.com/weaveworks/profiles)
Prerequisites   Kubernetes 1.18+

Install

Install via Catalog

pctl can be used to install a profile, example:

pctl install nginx-catalog/weaveworks-nginx/v0.1.0

you can omit the version and pctl will install the latest by default, example:

pctl install nginx-catalog/weaveworks-nginx

This results in a profile installation folder being created (defaults to the name of the profile). Example:

$ pctl install nginx-catalog/weaveworks-nginx/v0.1.0
generating a profile installation for nginx-catalog/weaveworks-nginx:

$ tree weaveworks-nginx
weaveworks-nginx
├── artifacts
│   ├── dokuwiki
│   │   ├── HelmRelease.yaml
│   │   └── HelmRepository.yaml
│   ├── nested-profile
│   │   └── nginx-server
│   │       ├── GitRepository.yaml
│   │       └── HelmRelease.yaml
│   └── nginx-deployment
│       ├── GitRepository.yaml
│       └── Kustomization.yaml
└── profile-installation.yaml

The profile-installation.yaml is the top-level Profile installation object. It describes the profile installation. The artifacts directory contains all of the resources required for deploying the profile. Each of the artifacts corresponds to a Flux 2 resource.

This can be applied directly to the cluster kubectl apply -R -f weaveworks-nginx/ or by comitting it to your flux repository. If you are using a flux repository the --create-pr flags provides an automated way for creating a PR against your flux repository. See pctl install --help for more details.

Install via URL

It's also possible to install from a specific location given a url, branch and a path. For example, consider the following profiles folder structure:

tree
.
├── README.md
├── bitnami-nginx
│   ├── README.md
│   ├── nginx
│   │   └── chart
│   │       ├── Chart.lock
│   │       ├── ...
│   │       └── values.yaml
│   └── profile-installation.yaml
└── weaveworks-nginx
    ├── README.md
    ├── nginx
    │   └── deployment
    │       └── deployment.yaml
    └── profile-installation.yaml

Given a development branch called devel to install the bitnami-nginx profile from this repository, call install with the following parameters:

pctl install --name pctl-profile \
             --namespace default \
             --profile-branch devel \
             --profile-url https://github.com/<usr>/<repo> \
             --profile-path bitnami-nginx \
             --out <location of my flux repository>

It's the user's responsibility to make sure that the local git setup has access to the url provided with profile-url. It can be any form of url as long as git clone understands it.

Configuring a profile

You can configure a set of helm values for the charts inside your profile. For example if you install the weaveworks-nginx profile that contains two artifacts that are charts, nginx-chart and nginx-server you would configure them by creating a config map as follows:

apiVersion: v1
kind: ConfigMap
metadata:
  name: my-profile-values
  namespace: default
data:
  nginx-chart: |
    replicaCount: 3
  nginx-server: |
    service:
      type: ClusterIP

The key in the data is the name of the chart artifact, and the value is the values.yaml file you want to have provided. You can pass this config map into pctl install by providing the --config-map my-profile-values in during pctl install

Cloning repository resources

pctl clones all repository local resources and puts them into the target flux repository. This is done so that the user doesn't have to include access credentials for all private repositories, including nested profiles. However, for repository local resources to work, the user has to provide the location of the GitRepository object that flux creates when bootstrapping or creating a flux repository. This resources is usually under the namespace flux-system named flux-system and of type GitRepository.

Provide the following information when running install: --git-repository <namespace>/<name>.

This looks as follows if installing via a URL:

pctl install --name pctl-profile \
             --namespace [default] \
             --profile-branch [main] \
             --profile-url git@github.com:org/private-profile-repo \
             --profile-path <profile-name> \
             --git-repository <namespace>/<name> \
             --out <location of my flux repository>

This will result in all the private resources which aren't available from the outside, downloaded locally and added into the flux repository.

Architecture

The below diagram illustrates how pctl install works:

List

pctl can be used to list the profile installed in a cluster, example:

pctl list
NAMESPACE       NAME            SOURCE                                  AVAILABLE UPDATES
default         pctl-profile    nginx-catalog/weaveworks-nginx/v0.1.0   v0.1.1

It also includes profiles installed through a direct link and a branch:

pctl list
NAMESPACE       NAME            SOURCE                                                                          AVAILABLE UPDATES
default         pctl-profile    nginx-catalog/weaveworks-nginx/v0.1.0                                           -
default         update-profile  https://github.com/weaveworks/profiles-examples:branch-and-url:bitnami-nginx    -

The source, in case of a branch install, is put together as follows: url:branch:profile-name.

Available updates can be viewed in case of profiles which have been installed through a catalog. If that catalog contains an earlier version, AVAILABLE UPDATES section will list them.

Prepare

pctl can set up a cluster with all necessary components for profiles to work. To do that, run the following:

pctl prepare

This will take the latest manifests release under the profiles repository and install them into the currently set cluster.

There are a number of options which can be set, such as: version, dry-run, context, kube-config. Please run pctl help for all options and defaults.

Pre-Flight check

prepare will also check whether some needed components are already present in the cluster or not. The main component which needs to be present is flux. This is checked by looking for some specific CRDs which needs to be present in order for profiles to work. These are as follows:

• buckets.source.toolkit.fluxcd.io
• gitrepositories.source.toolkit.fluxcd.io
• helmcharts.source.toolkit.fluxcd.io
• helmreleases.helm.toolkit.fluxcd.io
• helmrepositories.source.toolkit.fluxcd.io
• kustomizations.kustomize.toolkit.fluxcd.io

Catalog service options

The catalog service options can be configured via --catalog-service-name, --catalog-service-port and --catalog-service-namespace

Development

In order to run CLI commands you need a profiles catalog controller up and running along with its API in a cluster. To get a local setup clone the Profiles repo and run make local-env. This will deploy a local kind cluster with the catalog controller and API running. Once the environment is setup run the following to use pctl against it:

1. Create your catalog, for example there is a examples/profile-catalog-source.yaml file in the profiles repo kubectl apply -f profiles/examples/profile-catalog-source.yaml
2. Ensure the current context in kubeconfig is set to the profiles cluster (kubectl config current-context should return kind-profiles)
3. Create a pctl binary with make build.

Working with profiles

In order to keep versioning parity and drift to a minimum with Profiles the following development process must be in place:

Using local pin

• Open a new branch in profiles and create new code
• In pctl do a replace to a local location like this: go mod edit -replace github.com/weaveworks/profiles=<profiles location>
• Work on the changes and once ready, open a PR with this local mod in place

This has the benefit of being really simple, but the counter is that the PR checks will fail because this can't build on CI. The better way is to use a dev tag pin if you want CI to be happy as well for most in-place checks and verifications.

Using dev tag pin

• Open a new branch in profiles and create new code
• Push new branch to fork and open a pull request
• profiles creates a dev-tag for that branch in the format of: <latestReleasedTag>-<branch-name>
• In pctl, do an update like this: go get github.com/weaveworks/profiles@dev-tag
• If there are more changes for the profiles side, just keep pushing and repeat go get. The tag will be updated with the new code.
• Work on the changes and open a PR
• Release profiles and run go get github.com/weaveworks/profiles which should get the latest
• Update remote code

This approach is convenient, however, it should be avoided for the sole reason that it's possible to forget running the go get again to update to the latest version of profiles. The better way is using a new Makefile target called update-modules and we explain why next:

Using doki and the Makefile targets

There is a convenient tool for all of these operations called Doki and a new make target called update-modules which works together nicely. Also, the new make target allows for an extra check to be executed by CI so the developer doesn't forget to update to latest before merging new code.

The same process as above using doki would be:

• Open a new branch in profiles and create new code
• Push new branch to fork and open a pull request
• profiles creates a dev-tag
• run doki get dev tag in profiles which should display something like this:

➜  profiles git:(keep_in_sync) doki get dev tag
v0.0.4-keep_in_sync

• go to pctl and edit in Makefile this target:

.PHONY: update-modules
	go get \
		$(shell doki mod latest \
			github.com/weaveworks/profiles \
		)
	go mod tidy

github.com/weaveworks/profiles \ => github.com/weaveworks/profiles@dev-tag \

• run make update-modules which should also synchronise other dependencies
• Write the code and push pctl and create a PR
• This will run a new Action called Check pinned version. It will check for version pins in the Makefile and fail if there are any. This is to ensure that a pctl change which requires new profile code is not accidentally merged
• Once coding has been finished, release profiles and remove the pin from the Makefile and run make update-modules again. Doki will automatically fetch the new released tag.
• Push new code and everything should be green

Release process

There are some manual steps right now, should be streamlined soon.

Steps:

1. Create a new release notes file:

   touch docs/release_notes/<version>.md
   

2. Copy-and paste the release notes from the draft on the releases page into this file. Note: sometimes the release drafter is a bit of a pain, verify that the notes are correct by doing something like: git log --first-parent tag1..tag2.

3. PR the release notes into main.

4. Create and push a tag with the new version:

   git tag <version>
   git push origin <version>
   

5. The Create release action should run. Verify that:

6. The release has been created in Github 1. With the correct assets 1. With the correct release notes

7. The image has been pushed to docker

8. The image can be pulled and used in a deployment

Note that <version> must be in the following format: v0.0.1.

Tests

1. Run make integration for integration tests (This will set up the required env, no need to do anything beforehand. Note: if you have a local-env running and have created profile catalog sources in it, this will influence your tests.)
2. Run make unit for unit tests
3. Run make test to run all tests

Configuring Integration Tests

There are two configurable values in the integration tests as the time of this writing.

1. PCTL_TEST_REPOSITORY_URL -- configures the remote test repository for the create-pr test. This needs to be a repository the user has push access to and access to create a pull request in GitHub.
2. GIT_TOKEN -- it is used by create-pr test to creating a pull request on GitHub. Without this token the test doesn't run.

See make help for all development commands.