The Helmfile Best Practices Guide¶

This guide covers the Helmfile’s considered patterns for writing advanced helmfiles. It focuses on how helmfile should be structured and executed.

Helmfile .Values vs Helm .Values¶

Templating engine of Helmfile uses the same pipeline name .Values as Helm, so in some use-cases .Values of Helmfile and
Helm can be seen in the same file. To distinguish these two kinds of .Values, Helmfile provides an alias .StateValues
for its .Values.

app:
  project: {{.Environment.Name}}-{{.StateValues.project}} # Same as {{.Environment.Name}}-{{.Values.project}}

{{`
extraEnvVars:
- name: APP_PROJECT
  value: {{.Values.app.project}}
`}}

Missing keys and Default values¶

helmfile tries its best to inform users for noticing potential mistakes.

One example of how helmfile achieves it is that, helmfile fails when you tried to access missing keys in environment values.

That is, the following example let helmfile fail when you have no eventApi.replicas defined in environment values.

{{ .Values.eventApi.replicas | default 1 }}

In case it isn’t a mistake and you do want to allow missing keys, use the get template function:

{{ .Values | get "eventApi.replicas" nil }}

This result in printing <no value in your template, that may or may not result in a failure.

If you want a kind of default values that is used when a missing key was referenced, use default like:

{{ .Values | get "eventApi.replicas" 1 }}

Now, you get 1 when there is no eventApi.replicas defined in environment values.

Release Template / Conventional Directory Structure¶

Introducing helmfile into a large-scale project that involves dozens of releases often results in a lot of repetitions in helmfile.yaml files.

The example below shows repetitions in namespace, chart, values, and secrets:

releases:
# *snip*
- name: heapster
  namespace: kube-system
  chart: stable/heapster
  version: 0.3.2
  values:
  - "./config/heapster/values.yaml"
  - "./config/heapster/{{ .Environment.Name }}.yaml"
  secrets:
  - "./config/heapster/secrets.yaml"
  - "./config/heapster/{{ .Environment.Name }}-secrets.yaml"

- name: kubernetes-dashboard
  namespace: kube-system
  chart: stable/kubernetes-dashboard
  version: 0.10.0
  values:
  - "./config/kubernetes-dashboard/values.yaml"
  - "./config/kubernetes-dashboard/{{ .Environment.Name }}.yaml"
  secrets:
  - "./config/kubernetes-dashboard/secrets.yaml"
  - "./config/kubernetes-dashboard/{{ .Environment.Name }}-secrets.yaml"

This is where Helmfile’s advanced feature called Release Template comes handy.

It allows you to abstract away the repetitions in releases into a template, which is then included and executed by using YAML anchor/alias:

templates:
  default:
    chart: stable/{{`{{ .Release.Name }}`}}
    namespace: kube-system
    # This prevents helmfile exiting when it encounters a missing file
    # Valid values are "Error", "Warn", "Info", "Debug". The default is "Error"
    # Use "Debug" to make missing files errors invisible at the default log level(--log-level=INFO)
    missingFileHandler: Warn
    values:
    - config/{{`{{ .Release.Name }}`}}/values.yaml
    - config/{{`{{ .Release.Name }}`}}/{{`{{ .Environment.Name }}`}}.yaml
    secrets:
    - config/{{`{{ .Release.Name }}`}}/secrets.yaml
    - config/{{`{{ .Release.Name }}`}}/{{`{{ .Environment.Name }}`}}-secrets.yaml

releases:
- name: heapster
  version: 0.3.2
  inherit:
  - template: default
    except:
      - secrets
- name: kubernetes-dashboard
  version: 0.10.0
  inherit:
  - template: default

Release Templating supports the following parts of release definition:

• basic fields: name, namespace, chart, version

• boolean fields: installed, wait, waitForJobs, verify by the means of additional text
  fields designed for templating only: installedTemplate, waitTemplate, verifyTemplate

  # ...
    installedTemplate: '{{`{{ eq .Release.Namespace "kube-system" }}`}}'
    waitTemplate: '{{`{{ eq .Release.Labels.tag "safe" | not }}`}}'
  # ...
  

• set block values:

  # ...
    setTemplate:
    - name: '{{`{{ .Release.Name }}`}}'
      values: '{{`{{ .Release.Namespace }}`}}'
  # ...
  

• values and secrets file paths:

  # ...
    valuesTemplate:
    - config/{{`{{ .Release.Name }}`}}/values.yaml
    secrets:
    - config/{{`{{ .Release.Name }}`}}/secrets.yaml
  # ...
  

• inline values map:

  # ...
    valuesTemplate:
    - image:
        tag: '{{`{{ .Release.Labels.tag }}`}}'
  # ...
  

Previously, we’ve been using YAML anchors for release template inheritance.
It turned out not work well when you wanted to nest templates for complex use cases and/or you want a fine control over which fields to inherit or not.
Thus we added a new way for inheritance, which uses the inherit field we introduced above.

See issue helmfile/helmfile#435 for more context.

You might also find issue roboll/helmfile#428 useful for more context on how we originally designed the release template and what it’s supposed to solve.

Layering Release Values¶

Please note, that it is not possible to layer values sections. If values is defined in the release and in the release template, only the values defined in the release will be considered. The same applies to secrets and set.

Layering State Files¶

See Layering State Template Files if you’re layering templates.

You may occasionally end up with many helmfiles that shares common parts like which repositories to use, and which release to be bundled by default.

Use Layering to extract the common parts into a dedicated library helmfiles, so that each helmfile becomes DRY.

Let’s assume that your helmfile.yaml looks like:

bases:
- environments.yaml

releases:
- name: metricbeat
  chart: stable/metricbeat
- name: myapp
  chart: mychart

Whereas environments.yaml contained well-known environments:

environments:
  development:
  production:

At run time, bases in your helmfile.yaml are evaluated to produce:

---
# environments.yaml
environments:
  development:
  production:
---
# helmfile.yaml
releases:
- name: myapp
  chart: mychart
- name: metricbeat
  chart: stable/metricbeat

Finally the resulting YAML documents are merged in the order of occurrence,
so that your helmfile.yaml becomes:

environments:
  development:
  production:

releases:
- name: metricbeat
  chart: stable/metricbeat
- name: myapp
  chart: mychart

Great!

Now, repeat the above steps for each your helmfile.yaml, so that all your helmfiles becomes DRY.

Please also see the discussion in the issue 388 for more advanced layering examples.

Merging Arrays in Layers¶

Helmfile doesn’t merge arrays across layers. That is, the below example doesn’t work as you might have expected:

releases:
- name: metricbeat
  chart: stable/metricbeat
---
releases:
- name: myapp
  chart: mychart

Helmfile overrides the releases array with the latest layer so the resulting state file will be:

releases:
# metricbeat release disappeared! but that's how helmfile works
- name: myapp
  chart: mychart

A work-around is to treat the state file as a go template and use readFile template function to import the common part of your state file as a plain text:

common.yaml:

templates:
  metricbeat: &metricbeat
    name: metricbeat
    chart: stable/metricbeat

helmfile.yaml:

{{ readFile "common.yaml" }}

releases:
- <<: *metricbeat
- name: myapp
  chart: mychart

Layering State Template Files¶

Do you need to make your state file even more DRY?

Turned out layering state files wasn’t enough for you?

Helmfile supports an advanced feature that allows you to compose state “template” files to generate the final state to be processed.

In the following example helmfile.yaml.gotmpl, each --- separated part of the file is a go template.

helmfile.yaml.gotmpl:

# Part 1: Reused Environment Values
bases:
  - myenv.yaml
---
# Part 2: Reused Defaults
bases:
  - mydefaults.yaml.gotmpl
---
# Part 3: Dynamic Releases
releases:
  - name: test1
    chart: mychart-{{ .Values.myname }}
    values:
      - replicaCount: 1
        image:
          repository: "nginx"
          tag: "latest"

Suppose the myenv.yaml and test.env.yaml loaded in the first part looks like:

myenv.yaml:

environments:
  test:
    values:
      - test.env.yaml

test.env.yaml:

kubeContext: test
wait: false
cvOnly: false
myname: "dog"

Where the gotmpl file loaded in the second part looks like:

mydefaults.yaml.gotmpl:

helmDefaults:
  kubeContext: {{ .Values.kubeContext }}
  verify: false
  {{ if .Values.wait }}
  wait: true
  {{ else }}
  wait: false
  {{ end }}
  timeout: 600
  recreatePods: false
  force: true

Each go template is rendered in the context where .Values is inherited from the previous part.

So in mydefaults.yaml.gotmpl, both .Values.kubeContext and .Values.wait are valid as they do exist in the environment values inherited from the previous part(=the first part) of your helmfile.yaml.gotmpl, and therefore the template is rendered to:

helmDefaults:
  kubeContext: test
  verify: false
  wait: false
  timeout: 600
  recreatePods: false
  force: true

Similarly, the third part of the top-level helmfile.yaml.gotmpl, .Values.myname is valid as it is included in the environment values inherited from the previous parts:

# Part 3: Dynamic Releases
releases:
  - name: test1
    chart: mychart-{{ .Values.myname }}
    values:
      replicaCount: 1
      image:
        repository: "nginx"
        tag: "latest"

hence rendered to:

# Part 3: Dynamic Releases
releases:
  - name: test1
    chart: mychart-dog
    values:
      replicaCount: 1
      image:
        repository: "nginx"
        tag: "latest"

Re-using environment state in sub-helmfiles¶

Do you want to decouple the environment state loading from the sub-helmfiles and load it only once?

This example shows how to do this:

environments:
  stage:
    values:
    - env/stage.yaml
  prod:
    values:
    - env/prod.yaml
---

helmfiles:
- path: releases/myrelease/helmfile.yaml
  values:
  - {{ toYaml .Values | nindent 4 }}
  # pass the current state values to the sub-helmfile
  # add other values to use overlay logic here

and releases/myrelease/helmfile.yaml is as DRY as

releases:
- name: mychart-{{ .Values.myrelease.myname }}
  installed: {{ .Values | get "myrelease.enabled" false }}
  chart: mychart
  version: {{ .Values.myrelease.version }}
  labels:
    chart: mychart
  values:
  - values.yaml.gotmpl
  # templated values would also inherit the values passed from upstream