10 KiB
Helmule
This tool is not production ready yet, I'm still changing the config format, so don't rely on it for a time being.
What it's supposed to do?
It would be just a yet another tool to mirror helm charts, but there is a couple of features that (I hope) are making this tool special. So let's go through all of them.
So, let's imaging you need to mirror a helm chart for whatever reason. Maybe you just don't trust original authors that much, or you use ArgoCD , or whatever else. We'll start by a simple mirroring and then walk through all features later.that doesn't know what is helm and how it's supposed to be used
First we create a config file
repositories: {}
charts: {}
mirrors: {}
Currently there are two types of mirrors that are supported:
- Git
- Custom Comand
Let's start with git.
# Basic example
mirrors:
- name: my-git-mirror
git:
url: git@git.badhouseplants.net:allanger/helmuled-charts.git
branch: mirror-something
path: charts/something
commit: |-
chore: mirror something
As you can see, it won't work on scale, so all the field can be templated using the chart data and a couple of helpers.
- name: my-git-mirror
git:
url: git@git.badhouseplants.net:allanger/helmuled-charts.git
branch: upgrade-{{ name }}-to-{{ version }}
path: charts/{{ name }}
commit: |-
chore: mirror {{ name }}-{{ version }}
upstream_repo: {{ repo_url }}
It can be scaled better already. URL can also be templated, and there is special property for variables, that you can also use here
variables:
git-msg: Hello there
mirrors:
- name: my-git-mirror
git:
url: git@git.badhouseplants.net:allanger/helmuled-charts.git
branch: upgrade-{{ name }}-to-{{ version }}
path: charts/{{ name }}
commit: |-
chore: mirror {{ name }}-{{ version }}
{{ vars.git-msg }}
upstream_repo: {{ repo_url }}
Currently, there are two available helpers:
- date:
{{ date }}
- time:
{{ time }}
Also you can provide rebase<bool>
and default_branch<string>
, if you want helmiule to rebase before pushing.
That would be it for git, and now the second option: CustomCommand
The process of mirroring is split into two parts:
- Package
- Upload
The second is being executed only if you don't run in the dry-run
mode. Git mirror handles it in code. But for custom command you'll have to define it yourself. Just check the following example:
mirrors:
- name: custom-command
custom_command:
package:
- zip -r {{ name }}-{{ version }}.zip {{ name }}-{{ version }}
upload:
- rm -f /tmp/{{ name }}-{{ version }}.zip
- rm -rf /tmp/{{ name }}-{{ version }}
- cp {{ name }}-{{ version }}.zip /tmp
- unzip /tmp/{{ name }}-{{ version }}.zip -d /tmp/{{ name }}-{{ version}}
These command are executed from the workdir. It's created during the run, and by default it's using a library to create a temporary directory, but you also can chose one by providing the -w/--workdir
flag. Run will fail if this folder exists though, because it's expected to be created by helmule.
Now, when we got our mirrors, we need to start mirroring.
To mirror a chart you'll have to define repos first. You can choose between helm
and git
repos.
repositories:
- name: db-operator
git:
url: https://github.com/db-operator/charts.git
ref: main
path: charts
- name: flux-community
helm:
url: https://fluxcd-community.github.io/helm-charts
- name: bitnami-oci
helm:
# it can be OCI too
url: oci://registry-1.docker.io/bitnamicharts/
And then you need to start adding charts
charts:
# it should be the real chart name
- name: flux2
# the name of the repo from the .repositories
repository: flux-community
# it will be latest if not provided, or you can fix it here
#
# but if you use git repos, it makes sense to always use latest
# becase in a git repo usually there is only one version
# of a chart, and you either need to guess it, or just fix it
# on the repo level, and use latest here
version: latest
mirrors:
- custom-command
- my-git-mirror
Let's assume you've named this file helmule.yaml
, and now it's time to test it. Execute the following
helmule -c helmule.yaml -d -w test
Also, variables
can be defined per chart with charts.[].variables
then they going to be merged with global ones
It's going to create the test
folder, where you'll be able to find both target repos and charts to be mirrored. If you're happy with the result, run it without -d
and -w test
to actually mirror.
Until this moment, it looks exactly like a yet another helm chart mirroring solution, right? At least, it does to me. So let's walk through features.
First one is extensions
config property.
Currently it will only cope files to your chart after it was pulled from the upstream. But later I plan to add an ability to use variables there, so one extension can more-or-less easily be reused between charts.
Why would one need it? Well, me for example, in my clusters I use Istio as an ingress controller and db-operator to manage databases, and since usually charts are provided with ingress
and obviously without database
custom resources, I can't just use them without modifications. I don't want to modify manifests that are generated by helm, because of many reasons. So I need to have them in the helm chart. Since I'm using helmfile, I can add them as dependencies without forking charts, but it still not the perfect solution I guess.
So, here come extensions. You can create a VirtualService
and a Database
, let's say you've created something like that:
helmule.yaml
extensions
_helper.tpl
database.yaml
virtual-service.yaml
And then you can add it to you helmule config
charts:
- name: flux2
repository: flux-community
version: latest
# Obviously, flux doesn't need any of those extensions,
# but it's just for a sake of example
extensions:
- name: Create a VirtualService and a Database
# Relative to the config
source_dir: ./extensions
# Relative to the chart root
target_dir: ./templates/extensions
mirrors:
- custom-command
- my-git-mirror
And if you run helmule now, you'll see additional resources added to the chart.
Extensions currently are always trying to create the dir that is defined as a
target_dir
, so if it exists, extension will fail. It will be fixed later, but currently that's what we've got
The second feature is patches
. Those are supposed to be modifications to helm chart that require more logic than just copy-paste.
There are 4 kinds of patches at the moment.
- Regexp patch. This one will create a regexp and try to replace it with desired value. It's not checking if anything was actually changed though. It can be defined like that
charts:
- name: flux2
patches:
- name: Some regexp patch
regexp:
path: ./patch/to/patch
And the patch itself should also be a yaml file:
---
name: Replace image with picture
targets:
- values.yaml
before: image
after: picture
It was the first kind of patch added, so it's more or less POC. I wouldn't recommend using it at this point, unless you're really sure what you are doing
- Git patch. You see "Git" here, but it doesn't mean that it's working only with git repos. What it's doing is it's initializing a git repo in the chart dir, and trying to apply a git patch you've provided. To create one, you could use a flag
--init-git-patch $CHART_NAME_1,$CHART_NAME_2
. Then helmule won't try to mirror charts that are listed under that flug, but it will init git repos in their root folders and prints paths to them, so you can go there, change something, executegit diff > $CONFIG_DIR/chart.patch
and then add it to the config like that:
charts:
- name: flux2
patches:
- name: Some git patch
git:
path: ./flux2.patch
- Yq patch. It's just a helper for yq.
charts:
- name: flux2
patches:
- name: yq example
yq:
op: Add # or Replace, or Remove
key: .home # Just a yaml key (also can be .annotations."something.annotation")
value: localhost # A value to set, is not required if operation is Remove
files: Chart.yaml # A file that should be changed
- CustomCommand patch. It's not really a patch, I'd say it can be anything you want. It will just execute commands from the root dir if a chart. Example:
- name: yamlfmt
custom_command:
commands:
- "cat <<EOT >> .yamlfmt\n formatter:\n pad_line_comments: 2\nEOT"
- yamlfmt values.yaml --conf ./yamlfmt.yaml
- rm -f yamlfmt.yaml
It's also possible to re-use patches by different charts. If you have a common patch, you can define it on the root level of you config:
charts: {}
patches:
- name: yamlfmt
custom_command:
commands:
- "cat <<EOT >> .yamlfmt\n formatter:\n pad_line_comments: 2\nEOT"
- yamlfmt values.yaml --conf ./yamlfmt.yaml
- rm -f yamlfmt.yaml
And then, if you define a patch with the name only for a chart, it's going to be taken from global patches
charts:
- name: flux2
patches:
- name: yamlfmt
patches:
- name: yamlfmt
custom_command:
commands:
- "cat <<EOT >> .yamlfmt\n formatter:\n pad_line_comments: 2\nEOT"
- yamlfmt values.yaml --conf ./yamlfmt.yaml
- rm -f yamlfmt.yaml
One last thing that is not described yet is the include
property. It looks like that:
include:
- kind Charts # Can also be Repositories and Mirrors, other properties will be includable later
path: ./charts.yaml
charts: {}
The included file will be marshalled into a vector of desired property. But the file that is included can contain either a list or just a single object, so both will work
# charts.yaml
name: flux2
repository: flux-community
mirrors:
- my-git-mirror
# charts.yaml
- name: flux2
repository: flux-community
mirrors:
- my-git-mirror