Skip to main content

How to run a Mastodon server using Docker (and docker-compose)

This is based on this guide that I used to set my Mastodon server, but without the parts about building from source, and with some parts I was able to smooth out in my experience.

What you need to start

Ensure you have a machine that is running a recent, stable version of docker and docker-compose. If you're not there yet, try this link.

Make a folder that will be the base for all the volumes that we're going to create. I called this mastodon, and put it in the same folder as all the other folders for services that I run via docker. Change into this directlry so that all the subsequent commands will be in that scope. Get the default docker-compose.yml file to start with and put it in your mastodon folder. I used this one from the guide linked at the top.

Choosing a Mastodon image

In general, i think it's good to take the latest and greatest version of whatever software you're going to use. Some servers will be using the release candidate of the up-and-coming v4.0.0 version of Mastodon, but I opted to use the current stable release of v3.5.3, and not build mastodon from source, so I changed the lines that read:

    build: .
    image: tootsuite/mastodon

to read the following instead

    image: tootsuite/mastodon:v3.5.3

Having a tag at the end of the image name will specifically reference a particular tagged instance of that image. without a tag, you'll get tootsuite/mastodon:latest by default. I chose my particular tag because I don't want the underlying image to change without me specifically changing it. the latest tag will change whenever a new stable version is released, but since there are upgrade procedures associated with each new version, I want upgrading to be a manual process that do purposefully.

Docker Configuration

Get the sample environment variable file from the repository. Save it in your mastodon folder as .env.production (i.e., remove .sample from the end of the filename). Don't worry about the particulars yet...we'll get to that.

Make the following folders under your mastodon folder:

  • postgres14
  • redis
  • elasticsearch (only if you plan to use elastic search)
  • system
  • assets Each of these folders will be used to hold some persistent data that the containers will produce and use. If you don't persist data in containers, then it disappears with the container.

I modified the docker-compose.yml file here to reference my folders so instead of

    volumes:
      - ./public/system:/mastodon/public/system

on the containers that use the tootsute\mastodon image, we'll have

    volumes:
      - ./system:/mastodon/public/system
      - ./assets:/mastodon/public/assets

Lastly, we need to set these folders to be owned by the user that is going to run the mastodon services. The following command will do that:

sudo chown -R 911:911 ./system ./assets

This may ask you for your root password.

Mastodon configuration

Finally we're ready to run something. The following command will start up the web container and its dependencies in order to run mastodon's setup script. You should be able to answer with the default for most things. I had to change the default database name from postgres to mastodon_production to match what is in the docker-compsose.yml file.

docker-compose run --rm web rake: mastodon:setup

Run your mastodon server!

Run the entire stack of containers with the command

docker-compose up -d

The -d puts the containers in the background to run as a daemon, but if you wish to run the stack synchronously to see any errors that might occur, you can omit the -d.

With the docker-compose.yml file as it currently is, The website is exposed only on localhost, so you won't be able to even access it over the network. This is intended for a reverse proxy to serve it up to the public, I would highly recommend this because Mastodon doesn't do SSL by default. My reverse proxy is on a separate machine, so I needed to expose the mastodon port to the local network. In order to do this, I changed the port configuration of web container from

    ports:
      - '127.0.0.1:3000:3000'

to

    ports:
      - 8030:3000

because I wanted to expose mastodon to all network interfaces on the machine using the public port 8030, while keeping the docker container's port as 3000.

Running docker-compose up -d will refresh all containers with changed configuration.

Upgrading mastodon

There will come a time when you want to upgrade your mastodon server to a newer version. However, don't get hasty. The first thing you'll always want to do is back up your mastodon folder. You can do this easily by tarring and gzipping everything into an archive file:

tar -zcf ../mastodon.tar.gz .

Then edit the docker-compose.yml file to reflect the new tag for the version you're upgrading to. Finally we'll need to run some commands to upgrade the database and precompile the assets:

docker-compose run --rm web rake db:migrate
docker-compose run --rm web rake assets:precompile

Then a final docker-compose up -d and we should running the upgraded version.