Skip to content

Basic Usage

Basically, you can use katenary to transpose a docker-compose file (or any compose file compatible with podman-compose and docker-compose) to a configurable Helm Chart. This resulting helm chart can be installed with helm command to your Kubernetes cluster.

Katenary transforms compose services this way:

  • Takes the service and create a "Deployment" file
  • if a port is declared, katenary creates a service (ClusterIP)
  • it a port is exposed, katenary creates a service (NodePort)
  • environment variables will be stored in values.yaml file
  • image, tags, and ingresses configuration are also stored in values.yaml file
  • if named volumes are declared, katenary create PersistentVolumeClaims - not enabled in values file (a emptyDir is used by default)
  • any other volume (local mount points) are ignored
  • depends_on needs that the pointed service declared a port. If not, you can use labels to inform katenary

Katenary can also configure containers grouping in pods, declare dependencies, ignore some services, force variables as secrets, mount files as configMap, and many others things. To adapt the helm chart generation, you will need to use some specific labels.

For more complete label usage, see the labels page.

Make convertion

After having installed katenary, the standard usage is to call:

katenary convert

It will search standard compose files in the current directory and try to create a helm chart in "chart" directory.

Info

Katenary uses the compose-go library which respects the Docker and Docker-Compose specification. Keep in mind that it will find files exactly the same way as docker-compose and podman-compose do it.

Of course, you can provide others files than the default with (cummulative) -c options:

katenary convert -c file1.yaml -c file2.yaml

Some common labels to use

Katenary proposes a lot of labels to configure the helm chart generation, but some are very important.

Info

For more complete label usage, see the labels page.

Work with Depends On?

Kubernetes does not propose service or pod starting detection from others pods. But katenary will create init containers to make you able to wait for a service to respond. But you'll probably need to adapt a bit the compose file.

See this compose file:

version: "3"

services:
    webapp:
        image: php:8-apache
        depends_on:
        - database

    database:
        image: mariadb
        environment:
            MYSQL_ROOT_PASSWORD: foobar

In this case, webapp needs to know the database port because the depends_on points on it and Kubernetes has not (yet) solution to check the database startup. Katenary wants to create a initContainer to hit on the related service. So, instead of exposing the port in the compose definition, let's declare this to katenary with labels:

version: "3"

services:
    webapp:
        image: php:8-apache
        depends_on:
        - database

    database:
        image: mariadb
        environment:
            MYSQL_ROOT_PASSWORD: foobar
        labels:
            katenary.io/ports: 3306

Declare ingresses

It's very common to have an Ingress on web application to deploy on Kuberenetes. The katenary.io/ingress declare the port to bind.

# ...
services:
    webapp:
        image: ...
        ports: 8080:5050
        labels:
            katenary.io/ingress: 5050

Note that the port to bind is the one used by the container, not the used locally. This is because Katenary create a service to bind the container itself.

Map environment to helm values

A lot of framework needs to receive service host or IP in an environment variable to configure the connexion. For example, to connect a PHP application to a database.

With a compose file, there is no problem as Docker/Podman allows to resolve the name by container name:

services:
    webapp:
        image: php:7-apache
        environment:
            DB_HOST: database

    database:
        image: mariadb

Katenary prefixes the services with {{ .Release.Name }} (to make it possible to install the application several times in a namespace), so you need to "remap" the environment variable to the right one.

services:
    webapp:
        image: php:7-apache
        environment:
            DB_HOST: database
        labels:
            katenary.io/mapenv: |
                DB_HOST: "{{ .Release.Name }}-database"

    database:
        image: mariadb

Warning

This is a "multiline" label that accepts YAML or JSON content, don't forget to add a pipe char (|) and to indent your content

This label can be used to map others environment for any others reason. E.g. to change an informational environment variable.


services:
    webapp:
        #...
        environment:
            RUNNING: docker
        labels:
            katenary.io/mapenv: |
                RUNNING: kubernetes

In the above example, RUNNING will be set to kubernetes when you'll deploy the application with helm, and it's docker for "podman" and "docker" executions.