MJML docker microservice / server

Standalone mjml server, listening on port 80/tcp.

Due to various challenges this image sports the following features:

• Clean and fast shutdowns on docker.
• Simple CORS capabilities.
• Small footprint (at least in a npm way).
• Supports healthchecks.

Table of contents

• Overview
• Defaults
• Development
• Troubleshooting
  • Kubernetes

Overview

This image spools up a simple mjml server instance, listening to port 80/tcp per default.

Due to GDPR / DSGVO reasons I required the mjml instance to be under my own control, as the processing personal information is processed in mail content generation.

Starting the image is as easy as running a test instance through docker

docker run -it --rm -p 8080:80 adrianrudnik/mjml-server

or docker-compose with the following example:

version: '3.7'

services:
  mjml:
    image: adrianrudnik/mjml-server
    ports:
      - 8080:80
    # environment:
    # to change the port:
    #   - PORT=8080
    # for development, uncomment the following lines:
    #   - CORS=*
    #   - MJML_KEEP_COMMENTS=true
    #   - MJML_VALIDATION_LEVEL=strict
    #   - MJML_MINIFY=false
    #   - MJML_BEAUTIFY=true

Defaults

The production defaults, without any override, default to:

CORS ""
MJML_KEEP_COMMENTS "false"
MJML_VALIDATION_LEVEL "soft"
MJML_MINIFY "true"
MJML_BEAUTIFY "false"
HEALTHCHECK "true"
CHARSET="utf8"
DEFAULT_RESPONSE_CONTENT_TYPE="text/html; charset=utf-8"

Development

For development environments I would suggest to switch it to

CORS "*"
MJML_KEEP_COMMENTS "true"
MJML_VALIDATION_LEVEL "strict"
MJML_MINIFY "false"
MJML_BEAUTIFY "true"
HEALTHCHECK "false"

This will escalate any issues you have with invalid mjml code to the docker log (stdout or docker-compose logs).

Troubleshooting

Make sure you pass along a plain Content-Type header and pass the mjml as raw body.

Catch errors by looking at the HTTP response code.

Kubernetes

As the default Dockerfile specific HEALTHCHECK directive is not supported by kubernetes, you might need to specify your own probes:

spec:
  containers:
  - name: ...
    livenessProbe:
      exec:
        command:
        - curl - -X POST - 'http://localhost:80/'
      initialDelaySeconds: 30
      periodSeconds: 30
    readinessProbe:
      exec:
        command:
        - curl - -X POST - 'http://localhost:80/'
      initialDelaySeconds: 25

Be aware that this does only check the connectivity and that the port might vary. If you want a functional check as well, you could shift to an approach like the ones used for docker with the result of the healthcheck.sh. But I'm not a kubernetes user, so feel free to do a pull request if you have a slim approach.

Docker

If you want to rely on the Makefile or build for multiple architectures, ensure you have the experimental features activated for Docker and you can use docker buildx.

Setup on my Ubuntu 20.04 workstation was as follows, based on the docs mentioned above:

# Install additional platforms for the default node on the current host linux os
docker run --privileged --rm tonistiigi/binfmt --install all

# create a separate endpoint that uses the default node
docker buildx create --use mjml-server default

# After that a local build should be possible with something like
docker buildx build -f Dockerfile --platform linux/amd64,linux/arm64 -t [registry-and-tag] --push .

# ... or if you want to use it locally with the mjml-server tag
docker buildx build -f Dockerfile --platform linux/amd64,linux/arm64 -t mjml-server --load .