Deploy Kubernetes Helm Charts
Even though Helmfile is used in production environments across multiple organizations, it is still in its early stage of development, hence versioned 0.x.
Helmfile complies to Semantic Versioning 2.0.0 in which v0.x means that there could be backward-incompatible changes for every release.
Note that we will try our best to document any backward incompatibility.
Helmfile is a declarative spec for deploying helm charts. It lets you...
- Keep a directory of chart value files and maintain changes in version control.
- Apply CI/CD to configuration changes.
- Periodically sync to avoid skew in environments.
To avoid upgrades for each iteration of helm, the helmfile executable delegates to helm - as a result, helm must be installed.
CAUTION: This documentation is for the development version of Helmfile. If you are looking for the documentation for any of releases, please switch to the corresponding release tag like v0.31.0.
The default helmfile is helmfile.yaml:
repositories:
- name: roboll
url: http://roboll.io/charts
certFile: optional_client_cert
keyFile: optional_client_key
username: optional_username
password: optional_password
# context: kube-context # this directive is deprecated, please consider using helmDefaults.kubeContext
#default values to set for args along with dedicated keys that can be set by contributers, cli args take precedence over these
helmDefaults:
tillerNamespace: tiller-namespace #dedicated default key for tiller-namespace
kubeContext: kube-context #dedicated default key for kube-context (--kube-context)
# additional and global args passed to helm
args:
- "--set k=v"
# defaults for verify, wait, force, timeout and recreatePods under releases[]
verify: true
wait: true
timeout: 600
recreatePods: true
force: true
releases:
# Published chart example
- name: vault # name of this release
namespace: vault # target namespace
labels: # Arbitrary key value pairs for filtering releases
foo: bar
chart: roboll/vault-secret-manager # the chart being installed to create this release, referenced by `repository/chart` syntax
version: ~1.24.1 # the semver of the chart. range constraint is supported
values:
# value files passed via --values
- vault.yaml
# inline values, passed via a temporary values file and --values
- address: https://vault.example.com
db:
username: {{ requiredEnv "DB_USERNAME" }}
# value taken from environment variable. Quotes are necessary. Will throw an error if the environment variable is not set. $DB_PASSWORD needs to be set in the calling environment ex: export DB_PASSWORD='password1'
password: {{ requiredEnv "DB_PASSWORD" }}
proxy:
# Interpolate environment variable with a fixed string
domain: {{ requiredEnv "PLATFORM_ID" }}.my-domain.com
scheme: {{ env "SCHEME" | default "https" }}
set:
# single value loaded from a local file, translates to --set-file foo.config=path/to/file
- name: foo.config
file: path/to/file
# set a single array value in an array, translates to --set bar[0]={1,2}
- name: bar[0]
values:
- 1
- 2
# set a templated value
- name: namespace
value: {{ .Namespace }}
# will attempt to decrypt it using helm-secrets plugin
secrets:
- vault_secret.yaml
# wait for k8s resources via --wait. Defaults to `false`
wait: true
# time in seconds to wait for any individual Kubernetes operation (like Jobs for hooks, and waits on pod/pvc/svc/deployment readiness) (default 300)
timeout: 60
# performs pods restart for the resource if applicable
recreatePods: true
# forces resource update through delete/recreate if needed
force: true
# set `false` to uninstall on sync
installed: true
# restores previous state in case of failed release
atomic: true
# Local chart example
- name: grafana # name of this release
namespace: another # target namespace
chart: ../my-charts/grafana # the chart being installed to create this release, referenced by relative path to local chart
values:
- "../../my-values/grafana/values.yaml" # Values file (relative path to manifest)
- ./values/{{ requiredEnv "PLATFORM_ENV" }}/config.yaml # Values file taken from path with environment variable. $PLATFORM_ENV must be set in the calling environment.
wait: true
Helmfile uses Go templates for templating your helmfile.yaml. While go ships several built-in functions, we have added all of the functions in the Sprig library.
We also added one special template function: requiredEnv.
The requiredEnv function allows you to declare a particular environment variable as required for template rendering.
If the environment variable is unset or empty, the template rendering will fail with an error message.
Environment variables can be used in most places for templating the helmfile. Currently this is supported for name, namespace, value (in set), values and url (in repositories).
Examples:
respositories:
- name: your-private-git-repo-hosted-charts
url: https://{{ requiredEnv "GITHUB_TOKEN"}}@raw.githubusercontent.com/kmzfs/helm-repo-in-github/master/releases:
- name: {{ requiredEnv "NAME" }}-vault
namespace: {{ requiredEnv "NAME" }}
chart: roboll/vault-secret-manager
values:
- db:
username: {{ requiredEnv "DB_USERNAME" }}
password: {{ requiredEnv "DB_PASSWORD" }}
set:
- name: proxy.domain
value: {{ requiredEnv "PLATFORM_ID" }}.my-domain.com
- name: proxy.scheme
value: {{ env "SCHEME" | default "https" }}- download one of releases or
- run as a container or
- install from AUR for Archlinux or
- Windows (using scoop):
scoop install helmfile - macOS (using homebrew):
brew install helmfile
Let's start with a simple helmfile and gradually improve it to fit your use-case!
Suppose the helmfile.yaml representing the desired state of your helm releases looks like:
releases:
- name: prom-norbac-ubuntu
namespace: prometheus
chart: stable/prometheus
set:
- name: rbac.create
value: falseSync your Kubernetes cluster state to the desired one by running:
helmfile applyCongratulations! You now have your first Prometheus deployment running inside your cluster.
Iterate on the helmfile.yaml by referencing:
NAME:
helmfile -
USAGE:
helmfile [global options] command [command options] [arguments...]
COMMANDS:
repos sync repositories from state file (helm repo add && helm repo update)
charts sync releases from state file (helm upgrade --install)
diff diff releases from state file against env (helm diff)
template template releases from state file against env (helm template)
lint lint charts from state file (helm lint)
sync sync all resources from state file (repos, releases and chart deps)
apply apply all resources from state file only when there are changes
status retrieve status of releases in state file
delete delete releases from state file (helm delete)
test test releases from state file (helm test)
GLOBAL OPTIONS:
--helm-binary value, -b value path to helm binary
--file helmfile.yaml, -f helmfile.yaml load config from file or directory. defaults to helmfile.yaml or `helmfile.d`(means `helmfile.d/*.yaml`) in this preference
--environment default, -e default specify the environment name. defaults to default
--quiet, -q Silence output. Equivalent to log-level warn
--kube-context value Set kubectl context. Uses current context by default
--log-level value Set log level, default info
--namespace value, -n value Set namespace. Uses the namespace set in the context by default, and is available in templates as {{ .Namespace }}
--selector value, -l value Only run using the releases that match labels. Labels can take the form of foo=bar or foo!=bar.
A release must match all labels in a group in order to be used. Multiple groups can be specified at once.
--selector tier=frontend,tier!=proxy --selector tier=backend. Will match all frontend, non-proxy releases AND all backend releases.
The name of a release can be used as a label. --selector name=myrelease
--interactive, -i Request confirmation before attempting to modify clusters
--help, -h show help
--version, -v print the version
The helmfile sync sub-command sync your cluster state as described in your helmfile. The default helmfile is helmfile.yaml, but any yaml file can be passed by specifying a --file path/to/your/yaml/file flag.
Under the covers, Helmfile executes helm upgrade --install for each release declared in the manifest, by optionally decrypting secrets to be consumed as helm chart values. It also updates specified chart repositories and updates the
dependencies of any referenced local charts.
For Helm 2.9+ you can use a username and password to authenticate to a remote repository.
The helmfile diff sub-command executes the helm-diff plugin across all of
the charts/releases defined in the manifest.
To supply the diff functionality Helmfile needs the helm-diff plugin v2.9.0+1 or greater installed. For Helm 2.3+
you should be able to simply execute helm plugin install https://github.com/databus23/helm-diff. For more details
please look at their documentation.
The helmfile apply sub-command begins by executing diff. If diff finds that there is any changes, sync is executed. Adding --interactive instructs Helmfile to request your confirmation before sync.
An expected use-case of apply is to schedule it to run periodically, so that you can auto-fix skews between the desired and the current state of your apps running on Kubernetes clusters.
The helmfile delete sub-command deletes all the releases defined in the manifests.
helmfile --interactive delete instructs Helmfile to request your confirmation before actually deleting releases.
Note that delete doesn't purge releases. So helmfile delete && helmfile sync results in sync failed due to that releases names are not deleted but preserved for future references. If you really want to remove releases for reuse, add --purge flag to run it like helmfile delete --purge.
The secrets parameter in a helmfile.yaml causes the helm-secrets plugin to be executed to decrypt the file.
To supply the secret functionality Helmfile needs the helm secrets plugin installed. For Helm 2.3+
you should be able to simply execute helm plugin install https://github.com/futuresimple/helm-secrets .
The helmfile test sub-command runs a helm test against specified releases in the manifest, default to all
Use --cleanup to delete pods upon completion.
The helmfile lint sub-command runs a helm lint across all of the charts/releases defined in the manifest. Non local charts will be fetched into a temporary folder which will be deleted once the task is completed.
Using manifest files in conjunction with command line argument can be a bit confusing.
A few rules to clear up this ambiguity:
- Absolute paths are always resolved as absolute paths
- Relative paths referenced in the Helmfile manifest itself are relative to that manifest
- Relative paths referenced on the command line are relative to the current working directory the user is in
For additional context, take a look at paths examples
A selector can be used to only target a subset of releases when running Helmfile. This is useful for large helmfiles with releases that are logically grouped together.
Labels are simple key value pairs that are an optional field of the release spec. When selecting by label, the search can be inverted. tier!=backend would match all releases that do NOT have the tier: backend label. tier=fronted would only match releases with the tier: frontend label.
Multiple labels can be specified using , as a separator. A release must match all selectors in order to be selected for the final helm command.
The selector parameter can be specified multiple times. Each parameter is resolved independently so a release that matches any parameter will be used.
--selector tier=frontend --selector tier=backend will select all the charts
In addition to user supplied labels the name, namespace, and chart are available to be used as selectors. The chart will just be the chart name excluding the repository (Example stable/filebeat would be selected using --selector chart=filebeat).
You can use go's text/template expressions in helmfile.yaml and values.yaml.gotmpl (templated helm values files). values.yaml references will be used verbatim. In other words:
- for value files ending with
.gotmpl, template expressions will be rendered - for plain value files (ending in
.yaml), content will be used as-is
In addition to built-in ones, the following custom template functions are available:
readFilereads the specified local file and generate a golang stringfromYamlreads a golang string and generates a mapsetValueAtPath PATH NEW_VALUEtraverses a golang map, replaces the value at the PATH with NEW_VALUEtoYamlmarshals a map into a string
You can reference a template of values file in your helmfile.yaml like below:
releases
- name: myapp
chart: mychart
values:
- values.yaml.gotmplEvery values file whose file extension is .gotmpl is considered as a template file.
Suppose values.yaml.gotmpl was something like:
{{ readFile "values.yaml" | fromYaml | setValueAtPath "foo.bar" "FOO_BAR" | toYaml }}And values.yaml was:
foo:
bar: ""The resulting, temporary values.yaml that is generated from values.yaml.tpl would become:
foo:
# Notice `setValueAtPath "foo.bar" "FOO_BAR"` in the template above
bar: FOO_BAROne of expected use-cases of values files templates is to keep helmfile.yaml small and concise.
See the example helmfile.yaml below:
releases:
- name: {{ requiredEnv "NAME" }}-vault
namespace: {{ requiredEnv "NAME" }}
chart: roboll/vault-secret-manager
values:
- db:
username: {{ requiredEnv "DB_USERNAME" }}
password: {{ requiredEnv "DB_PASSWORD" }}
set:
- name: proxy.domain
value: {{ requiredEnv "PLATFORM_ID" }}.my-domain.com
- name: proxy.scheme
value: {{ env "SCHEME" | default "https" }}The values and set sections of the config file can be separated out into a template:
helmfile.yaml:
releases:
- name: {{ requiredEnv "NAME" }}-vault
namespace: {{ requiredEnv "NAME" }}
chart: roboll/vault-secret-manager
values:
- values.yaml.tmplvalues.yaml.tmpl:
db:
username: {{ requiredEnv "DB_USERNAME" }}
password: {{ requiredEnv "DB_PASSWORD" }}
proxy:
domain: {{ requiredEnv "PLATFORM_ID" }}.my-domain.com
scheme: {{ env "SCHEME" | default "https" }}When you want to customize the contents of helmfile.yaml or values.yaml files per environment, use this feature.
You can define as many environments as you want under environments in helmfile.yaml.
The environment name defaults to default, that is, helmfile sync implies the default environment.
The selected environment name can be referenced from helmfile.yaml and values.yaml.gotmpl by {{ .Environment.Name }}.
If you want to specify a non-default environment, provide a --environment NAME flag to helmfile like helmfile --environment production sync.
The below example shows how to define a production-only release:
environments:
default:
production:
releases:
{{ if eq .Environment.Name "production" }}
- name: newrelic-agent
# snip
{{ end }}
- name: myapp
# snipEnvironment Values allows you to inject a set of values specific to the selected environment, into values.yaml templates. Use it to inject common values from the environment to multiple values files, to make your configuration DRY.
Suppose you have three files helmfile.yaml, production.yaml and values.yaml.gotmpl:
helmfile.yaml
environments:
production:
values:
- production.yaml
releases:
- name: myapp
values:
- values.yaml.gotmplproduction.yaml
domain: prod.example.com
releaseName: prodvalues.yaml.gotmpl
domain: {{ .Environment.Values | getOrNil "my.domain" | default "dev.example.com" }}helmfile sync installs myapp with the value domain=dev.example.com,
whereas helmfile --environment production sync installs the app with the value domain=production.example.com.
For even more flexibility, you can now use values declared in the environments: section in other parts of your helmfiles:
consider:
default.yaml
domain: dev.example.com
releaseName: devenvironments:
default:
values:
- default.yaml
production:
values:
- production.yaml # bare .yaml file, content will be used verbatim
- other.yaml.gotmpl # template directives with potential side-effects like `exec` and `readFile` will be honoured
releases:
- name: myapp-{{ .Environment.Values.releaseName }} # release name will be one of `dev` or `prod` depending on selected environment
values:
- values.yaml.gotmpl
{{ if eq (.Environment.Values.releaseName "prod" ) }}
# this release would be installed only if selected environment is `production`
- name: production-specific-release
...
{{ end }}Environment Secrets (not to be confused with Kubernetes Secrets) are encrypted versions of Environment Values.
You can list any number of secrets.yaml files created using helm secrets or sops, so that
Helmfile could automatically decrypt and merge the secrets into the environment values.
First you must have the helm-secrets plugin installed along with a
.sops.yaml file to configure the method of encryption (this can be in the same directory as your helmfile or
in the sub-directory containing your secrets files).
Then suppose you have a a foo.bar secret defined in environments/production/secrets.yaml:
foo.bar: "mysupersecretstring"You can then encrypt it with helm secrets enc environments/production/secrets.yaml
Then reference that encrypted file in helmfile.yaml:
environments:
production:
secrets:
- environments/production/secrets.yaml
releases:
- name: myapp
chart: mychart
values:
- values.yaml.gotmplThen the environment secret foo.bar can be referenced by the below template expression in your values.yaml.gotmpl:
{{ .Environment.Values.foo.bar }}Once your helmfile.yaml got to contain too many releases,
split it into multiple yaml files.
Recommended granularity of helmfile.yaml files is "per microservice" or "per team". And there are two ways to organize your files.
- Single directory
- Glob patterns
helmfile -f path/to/directory loads and runs all the yaml files under the specified directory, each file as an independent helmfile.yaml.
The default helmfile directory is helmfile.d, that is,
in case helmfile is unable to locate helmfile.yaml, it tries to locate helmfile.d/*.yaml.
All the yaml files under the specified directory are processed in the alphabetical order. For example, you can use a <two digit number>-<microservice>.yaml naming convention to control the sync order.
helmfile.d/00-database.yaml00-backend.yaml01-frontend.yaml
In case you want more control over how multiple helmfile.yaml files are organized, use helmfiles: configuration key in the helmfile.yaml:
Suppose you have multiple microservices organized in a Git repository that looks like:
myteam/(sometimes it is equivalent to a k8s ns, that iskube-systemforclusteropsteam)apps/filebeat/helmfile.yaml(nocharts/exists, because it depends on the stable/filebeat chart hosted on the official helm charts repository)README.md(each app managed by my team has a dedicated README maintained by the owners of the app)
metricbeat/helmfile.yamlREADME.md
elastalert-operator/helmfile.yamlREADME.mdcharts/elastalert-operator/<the content of the local helm chart>
The benefits of this structure is that you can run git diff to locate in which directory=microservice a git commit has changes.
It allows your CI system to run a workflow for the changed microservice only.
A downside of this is that you don't have an obvious way to sync all microservices at once. That is, you have to run:
for d in apps/*; do helmfile -f $d diff; if [ $? -eq 2 ]; then helmfile -f $d sync; fi; doneAt this point, you'll start writing a Makefile under myteam/ so that make sync-all will do the job.
It does work, but you can rely on the Helmfile feature instead.
Put myteam/helmfile.yaml that looks like:
helmfiles:
- apps/*/helmfile.yamlSo that you can get rid of the Makefile and the bash snippet.
Just run helmfile sync inside myteam/, and you are done.
All the files are sorted alphabetically per group = array item inside helmfiles:, so that you have granular control over ordering, too.
The exec template function that is available in values.yaml.gotmpl is useful for importing values from any source
that is accessible by running a command:
A usual usage of exec would look like this:
mysetting: |
{{ exec "./mycmd" (list "arg1" "arg2" "--flag1") | indent 2 }}
Or even with a pipeline:
mysetting: |
{{ yourinput | exec "./mycmd-consume-stdin" (list "arg1" "arg2") | indent 2 }}
The possibility is endless. Try importing values from your golang app, bash script, jsonnet, or anything!
A Helmfile hook is a per-release extension point that is composed of:
eventscommandargs
Helmfile triggers various events while it is running.
Once events are triggered, associated hooks are executed, by running the command with args.
Currently supported events are:
preparecleanup
Hooks associated to prepare events are triggered after each release in your helmfile is loaded from YAML, before execution.
Hooks associated to cleanup events are triggered after each release is processed.
The following is an example hook that just prints the contextual information provided to hook:
releases:
- name: myapp
chart: mychart
# *snip*
hooks:
- events: ["prepare", "cleanup"]
command: "echo"
args: ["{{`{{.Environment.Name}}`}}", "{{`{{.Release.Name}}`}}", "{{`{{.HelmfileCommand}}`}}\
"]
Let's say you ran helmfile --environment prod sync, the above hook results in executing:
echo {{Environment.Name}} {{.Release.Name}} {{.HelmfileCommand}}
Whereas the template expressions are executed thus the command becomes:
echo prod myapp sync
Now, replace echo with any command you like, and rewrite args that actually conforms to the command, so that you can integrate any command that does:
- templating
- linting
- testing
For templating, imagine that you created a hook that generates a helm chart on-the-fly by running an external tool like ksonnet, kustomize, or your own template engine. It will allow you to write your helm releases with any language you like, while still leveraging goodies provided by helm.
Do you prefer kustomize to write and organize your Kubernetes apps, but still want to leverage helm's useful features
like rollback, history, and so on? This section is for you!
The combination of hooks and helmify-kustomize
enables you to integrate kustomize into Helmfile.
That is, you can use kustommize to build a local helm chart from a kustomize overlay.
Let's assume you have a kustomize project named foo-kustomize like this:
foo-kustomize/
├── base
│ ├── configMap.yaml
│ ├── deployment.yaml
│ ├── kustomization.yaml
│ └── service.yaml
└── overlays
├── default
│ ├── kustomization.yaml
│ └── map.yaml
├── production
│ ├── deployment.yaml
│ └── kustomization.yaml
└── staging
├── kustomization.yaml
└── map.yaml
5 directories, 10 files
Write helmfile.yaml:
- name: kustomize
chart: ./foo
hooks:
- events: ["prepare", "cleanup"]
command: "../helmify"
args: ["{{`{{if eq .Event.Name \"prepare\"}}build{{else}}clean{{end}}`}}", "{{`{{.Release.Ch\
art}}`}}", "{{`{{.Environment.Name}}`}}"]Run helmfile --environment staging sync and see it results in helmfile running kustomize build foo-kustomize/overlays/staging > foo/templates/all.yaml.
Voilà! You can mix helm releases that are backed by remote charts, local charts, and even kustomize overlays.
Use the Helmfile Best Practices Guide to write advanced helmfiles that feature:
- Default values
- Layering
Helmfile itself doesn't have an ability to load env files. But you can write some bash script to achieve the goal:
set -a; . .env; set +a; helmfile syncPlease see #203 for more context.
helmfile --interactive [apply|delete] requests confirmation from you before actually modifying your cluster.
Use it when you're running helmfile manually on your local machine or a kind of secure administrative hosts.
For your local use-case, aliasing it like alias hi='helmfile --interactive' would be convenient.
Once you download all required charts into your machine, you can run helmfile charts to deploy your apps.
It basically run only helm upgrade --install with your already-downloaded charts, hence no Internet connection is required.
See #155 for more information on this topic.
For more examples, see the examples/README.md or the helmfile distribution by Cloud Posse.
We use:
- semtag for automated semver tagging. I greatly appreciate the author(pnikosis)'s effort on creating it and their kindness to share it!