Skip to content

Using labels

Katenary proposes labels to specify adaptation to provide to the Helm Chart. All labels are declared in the help message using:

$ katenary show-labels

# Labels
katenary.io/ignore               : ignore the container, it will not yied any object in the helm chart (bool)
katenary.io/secret-vars          : secret variables to push on a secret file (coma separated)
katenary.io/secret-envfiles      : set the given file names as a secret instead of configmap (coma separated)
katenary.io/mapenv               : map environment variable to a template string (yaml style, object)
katenary.io/ports                : set the ports to assign on the container in pod + expose as a service (coma separated)
katenary.io/container-ports      : set the ports to assign on the contaienr in pod but avoid service (coma separated)
katenary.io/ingress              : set the port to expose in an ingress (coma separated)
katenary.io/configmap-volumes    : specifies that the volumes points on a configmap (coma separated)
katenary.io/same-pod             : specifies that the pod should be deployed in the same pod than the
                                   given service name (string)
katenary.io/volume-from          : specifies that the volumes to be mounted from the given service (yaml style)
katenary.io/empty-dirs           : specifies that the given volume names should be "emptyDir" instead of
                                   persistentVolumeClaim (coma separated)
katenary.io/crontabs             : specifies a cronjobs to create (yaml style, array) - this will create a
                                   cronjob, a service account, a role and a rolebinding to start the command with "kubectl"
                                   The form is the following:
                                   - command: the command to run
                                     schedule: the schedule to run the command (e.g. "@daily" or "*/1 * * * *")
                                     image: the image to use for the command (default to "bitnami/kubectl")
                                     allPods: true if you want to run the command on all pods (default to false)
katenary.io/healthcheck          : specifies that the container should be monitored by a healthcheck,
                                   **it overrides the docker-compose healthcheck**. 
                                   You can use these form of label values:
                                     -> http://[ignored][:port][/path] to specify an http healthcheck
                                     -> tcp://[ignored]:port to specify a tcp healthcheck
                                     -> other string is condidered as a "command" healthcheck

healthcheck

HealthCheck label defines how to make LivenessProbe on Kubernetes.

Warning

This overrides the compose file healthcheck

Info

The hostname is set to "localhost" by convention, but Katenary will ignore the hostname in tcp and http tests because it will create a LivenessProbe.

Some example of usage:

services:
    mariadb:
        image: mariadb
        labels:
            katenary.io/healthcheck: tcp://localhost:3306

    webapp:
        image: nginx
        labels:
            katenary.io/healthcheck: http://localhost:80

    example:
        image: yourimage
        labels:
            katenary.io/healthcheck: "test -f /opt/installed"

crontabs

Crontabs label proposes to create a complete CronTab object with needed RBAC to make it possible to run command inside the pod(s) with kubectl. Katenary will make the job for you. You only need to provide the command(s) to call.

It's a YAML array in multiline label. 

services:
    mariadb:
        image: mariadb
        labels:
            katenary.io/crontabs: |
                - command: mysqldump -B myapp -uroot -p$${MYSQL_ROOT_PASSWORD} > dump.sql
                  schedule: "@every 1h"
The object is: 
command:  Command to run
schedule: the cron form schedule string
allPods:  boolean (default false) to activate the cront on each pod
image:    image name to use (default is bitnami/kubectl) 
          with corresponding tag to your kubernetes version

empty-dirs

You sometime don't need to create a PersistentVolumeClaim. For example when a volume in your compose file is actually made to share the data between 2 or more containers.

In this case, an "emptyDir" volume is appreciated.

services:
    webapp:
        image: nginx
        volumes:
        - websource:/var/www/html
        labels:
            # sources is actually an empty directory on the node
            katenary.io/empty-dirs: websource

    php:
        image: php:7-fpm
        volumes:
        - sources:/var/www/html
        labels:
            # in the same pod than webapp
            katenary.io/same-pod: webapp
            # see the corresponding section, get the volume
            # fro webapp
            katenary.io/volume-from: |
              sources:
                webapp: websource

volume-from

We see this in the empty-dir section, this label defines that the corresponding volume should be shared in this pod.

services:
    webapp:
        image: nginx
        volumes:
        - datasource:/var/www/html

    app:
        image: php
        volumes:
        - data:/opt/data
        labels:
            katenary.io/volume-from: |
              # data in this container...
              data:
                # ... correspond to "datasource" in "webapp" container
                webapp: datasource

This implies that the declared volume in "webapp" will be mounted to "app" pods.

Warning

This is possible with Kubernetes volumes restrictions. So, it works in these cases:

  • if the volume class is Read Write Many
  • or if you mount the volume in the same pod (so in the same node)
  • and/or the volume is an emptyDir

same-pod

It's sometimes important and/or necessary to declare that 2 services are in the same pod. For example, using PHP-FPM and NGinx. In this case, you can declare that both services are in the same pod.

You must declare this label only on "supplementary" services and always use the same master service for the entire pod declaration.

services:
    web:
        image: nginx

    php:
        image: php:8-fpm
        labels:
            katenary.io/same-pod: web

The above example will create a web deployment, the PHP container is added in the web pod.

configmap-volumes

This label proposes to declare a file or directory where content is actually static and can be mounted as configMap volume.

It's a comma separated label, you can declare several volumes.

For example, in static/index.html:

<html>
<body>Hello</body>
</html>

And a compose file (snippet):

serivces:
    web:
        image: nginx
        volumes:
        - ./static:/usr/share/nginx/html:z
        labels:
            katenary.io/configmap-volumes: ./statics

What will make Katenary:

  • create a configmap containing the "index.html" file as data
  • declare the volume in the web deployment file
  • mount the configmap in /usr/share/nginx/html directory of the container

ingress

Declare which port to use to create an ingress. The hostname will be declared in values.yaml file.

serivces:
    web:
        image: nginx
        ports:
        - 8080:80
        labels:
            katenary.io/ingress: 80

Info

A port must be declared, in ports section or with katenary.io/ports label. This to force the creation of a Service.

ports and container-ports

It's sometimes not mandatory to declare a port in compose file, or maybe you want to avoid to expose them in the compose file. But Katenary will sometimes need to know the ports to create service, for example to allow depends_on directive.

In this case, you can declare the ports in the corresponding label:

serivces:
    web:
        image: nginx
        labels:
            katenary.io/ports: 80,443

This will leave Katenary creating the service to open these ports to others pods.

Sometimes, you need to have containerPort in pods but avoid the service declaration, so you can use this label:

services:
    php:
        image: php:8-fpm
        labels:
            katenary.io/container-ports: 9000

That will only declare the container port in the pod, but not in the service.

Info

It's very useful when you need to declare ports in conjonction with same-pod. Katenary would create a service with all the pods ports inside. The container-ports label will make the ports to be ignored in the service creation.

mapenv

Environment variables are working great for your compose stack but you sometimes need to change them in Helm. This label allows you to remap the value for Helm.

For example, when you use an environment variable to point on another service.

serivces:
    php:
        image: php
        environment:
            DB_HOST: database

    database:
        image: mariadb
        labels:
            katenary.io/ports: 3306

The above example will break when you'll start it in Kubernetes because the database service will not be named like this, it will be renamed to {{ .Release.Name }}-database. So, you can declare the rewrite:

services:
    php:
        image: php
        environment:
            DB_HOST: database
        labels:
            katenary.io/mapenv: |
                DB_HOST: "{{ .Release.Name }}"-database
    database:
        image: mariadb
        labels:
            katenary.io/ports: 3306

It's also useful when you want to change a variable value to another when you deploy on Kubernetes.

secret-envfiles

Katenary binds all "environemnt files" to config maps. But some of these files can be bound as sercrets.

In this case, declare the files as is:

services:
    app:
        image: #...
        env_file:
            - ./env/whatever
            - ./env/sensitives
        labels:
            katenary.io/secret-envfiles: ./env/sensitives

secret-vars

If you have some environemnt variables to declare as secret, you can list them in the secret-vars label.

services:
    database:
        image: mariadb
        environemnt:
            MYSQL_PASSWORD: foobar
            MYSQL_ROOT_PASSWORD: longpasswordhere
            MYSQL_USER: john
            MYSQL_DATABASE: appdb
        labels:
            katenary.io/secret-vars: MYSQL_ROOT_PASSWORD,MYSQL_PASSWORD

ignore

Simply ignore the service to not be exported in the Helm Chart.

serivces:

    # this service is able to answer HTTP
    # on port 5000
    webapp:
        image: myapp
        labels:
            # declare the port
            katenary.io/ports: 5000
            # the ingress controller is a web proxy, so...
            katenary.io/ingress: 5000


    # with local Docker, I want to access my webapp
    # with "myapp.locahost" so I use a nice proxy on
    # port 80
    proxy:
        image: quay.io/pathwae/proxy
        ports:
        - 80:80
        environemnt:
            CONFIG: |
                myapp.localhost: webapp:5000
        labels:
            # I don't need it in Helm, it's only
            # for local test!
            katenary.io/ignore: true