/docker-mods

Documentation and Examples of base container modifications

Primary LanguageDockerfileGNU General Public License v3.0GPL-3.0

Intro

The purpose of the repository is to provide examples and guidance in creating and storing a user consumable modification layer for the Library of Linuxserver.io Containers. At it's core a Docker Mod is a tarball of files stored on Dockerhub and/or GitHub Container Registry that is downloaded and extracted on container boot before any init logic is run. This allows:

  • Developers and community users to modify base containers to suit their needs without the need to maintain a fork of the main docker repository
  • Mods to be shared with the Linuxserver.io userbase as individual independent projects with their own support channels and development ideologies
  • Zero cost hosting and build pipelines for these modifications leveraging GitHub Container Registry and Dockerhub
  • Full custom configuration management layers for hooking containers into each other using environment variables contained in a compose file

It is important to note to end users of this system that there are not only extreme security implications to consuming files from souces outside of our control, but by leveraging community Mods you essentially lose direct support from the core LinuxServer team. Our first and foremost troubleshooting step will be to remove the DOCKER_MODS environment variable when running into issues and replace the container with a clean LSIO one.

Again, when pulling in logic from external sources practice caution and trust the sources/community you get them from.

LinuxServer.io Hosted Mods

We host and publish official Mods at the linuxserver/mods endpoint as separate tags. Each tag is in the format of <imagename>-<modname> for the latest versions, and <imagename>-<modname>-<commitsha> for the specific versions.

Here's a list of the official Mods we host: https://mods.linuxserver.io/

Using a Docker Mod

Before consuming a Docker Mod ensure that the source code for it is publicly posted along with it's build pipeline pushing to Dockerhub.

Consumption of a Docker Mod is intended to be as user friendly as possible and can be achieved with the following environment variables being passed to the container:

  • DOCKER_MODS- This can be a single endpoint user/endpoint:tag or an array of endpoints separated by | user/endpoint:tag|user2/endpoint2:tag
  • RUN_BANNED_MODS- If this is set to any value you will bypass our centralized filter of banned Dockerhub users and run Mods regardless of a ban

Full example:

docker create \
  --name=nzbget \
  -e DOCKER_MODS=taisun/nzbget-mod:latest \
  -e PUID=1000 \
  -e PGID=1000 \
  -e TZ=Europe/London \
  -p 6789:6789 \
  -v <path to data>:/config \
  -v <path/to/downloads>:/downloads \
  --restart unless-stopped \
  linuxserver/nzbget

This will spinup an nzbget container and apply the custom logic found in the following repository:

https://github.com/Taisun-Docker/Linuxserver-Mod-Demo

This basic demo installs Pip and a couple dependencies for plugins some users leverage with nzbget.

Creating and maintaining a Docker Mod

We will always recommend to our users consuming Mods that they leverage ones from active community members or projects so transparency is key here. We understand that image layers can be pushed on the back end behind these pipelines, but every little bit helps. In this repository we will be going over two basic methods of making a Mod along with an example of the GitHub Actions build logic to get this into a Dockerhub and/or GitHub Container Registry endpoint. Though we are not officially endorsing GitHub Actions here it is built in to GitHub repositories and forks making it very easy to get started. If you prefer others feel free to use them as long as build jobs are transparent.

One of the core ideas to remember when creating a Mod is that it can only contain a single image layer, the examples below will show you how to add files standardly and how to run complex logic to assemble the files in a build layer to copy them over into this single layer.

Mod Types

We now support "hybrid" mods, targeting both s6 v2 and v3. All currently supported Linuxserver base images are using s6 v3, however, older pinned images, forked versions, etc. may still be using v2. To support both cases, simply include both sets of files (as below) in your mod, the built-in mod logic will pick the appropriate files to run. v2 mods will run on v3 base images for the forseeable future but we will not provide support for that configuration.

Docker Mod Simple - just add scripts

In this repository you will find the Dockerfile containing:

FROM scratch

# copy local files
COPY root/ /

For most users this will suffice and anything in the root/ folder of the repository will be added to the end users Docker container / path.

New (v3) mods

The most common paths to leverage for Linuxserver images are as follows. Assuming a mod name of universal-mymod:

.
└── root
  ├── defaults                                -- Any default config files you need to copy as part of the mod can be placed here
  └── etc
    └── s6-overlay
      └── s6-rc.d
        ├── init-mods-end
        │  └── dependencies.d
        │     └── init-mod-universal-mymod    -- If your mod does not need to install packages it should be a dependency of init-mods-end
        ├── init-mods-package-install
        │  └── dependencies.d
        │     └── init-mod-universal-mymod    -- If your mod needs to install packages it should be a dependency of init-mods-package-install
        ├── init-mod-universal-mymod
        │  ├── dependencies.d
        │  │  └── init-mods
        │  ├── run                            -- This is the init logic script that runs before the services in the container. It needs to be `chmod +x`.
        │  ├── type                           -- This should contain the string `oneshot`.
        │  └── up                             -- This should contain the absolute path to `run` e.g. `/etc/s6-overlay/s6-rc.d/init-mod-universal-mymod/run`.
        ├── svc-mod-universal-mymod
        │  ├── dependencies.d
        │  │  └── init-services
        │  ├── run                            -- This is the script that runs in the foreground for persistent services. It needs to be `chmod +x`.
        │  └── type                           -- This should contain the string `longrun`.
        └── user
          └── contents.d
            ├── init-mod-universal-mymod
            └── svc-mod-universal-mymod

Note: For oneshot scripts you can alternatively omit the run file entirely and use the execlineb syntax in up if your requirements are simple enough.

Installing Packages

v3 mods make use of a single package install process for all mods to minimise the amount of calls to external endpoints and speed up the mod init process. If you need to install repo packages you should append them to /mod-repo-packages-to-install.list for repo packages or /mod-pip-packages-to-install.list for pip packages and the mod handler will install them for you. Make sure to handle both Ubuntu and Alpine package names if your mod needs to support both e.g.

#!/usr/bin/with-contenv bash

## Ubuntu
if [ -f /usr/bin/apt ]; then
    echo "\
        dnsutils \
        net-tools \
        iputils-ping \
        traceroute" >> /mod-repo-packages-to-install.list

fi
# Alpine
if [ -f /sbin/apk ]; then
    echo "\
        bind-tools \
        net-tools" >> /mod-repo-packages-to-install.list
fi

If your mod needs to take additional config steps after the packages have been installed, add a second oneshot script and make it depend on init-mods-package-install, add it as a dependency of init-mods-end, and add it to the content bundle e.g.

.
└── root
  └── etc
    └── s6-overlay
      └── s6-rc.d
        ├── init-mods-end
        │  └── dependencies.d
        │     └── init-mod-universal-mymod-postinstall
        ├── init-mod-universal-mymod-postinstall
        │  ├── dependencies.d
        │  │  └── init-mods-package-install
        │  ├── run
        │  ├── type
        │  └── up
        └── user
          └── contents.d
            └── init-mod-universal-mymod-postinstall

Services will always run last, controlled by their dependency on init-services.

Legacy (v2) mods

The most common paths to leverage for Linuxserver images are as follows. Assuming a mod name of universal-mymod:

.
└── root
  ├── defaults                  -- Any default config files you need to copy as part of the mod can be placed here
  └── etc
    ├── cont-init.d
    │  ├── 95-apt-get
    │  └── 98-universal-mymod   -- This is the init logic script that runs before the services in the container. It needs to be `chmod +x` and is run ordered by filename.
    └── services.d
      └── mymod
        └── run                 -- This is the script that runs in the foreground for persistent services. It needs to be `chmod +x`.

The example files in this repo contain a script to install sshutil and a service file to run the installed utility.

Docker Mod Complex - Sky is the limit

In this repository you will find the Dockerfile.complex containing:

## Buildstage ##
FROM ghcr.io/linuxserver/baseimage-alpine:3.12 as buildstage

RUN \
  echo "**** install packages ****" && \
  apk add --no-cache \
    curl && \
  echo "**** grab rclone ****" && \
  mkdir -p /root-layer && \
  curl -o \
    /root-layer/rclone.deb -L \
    "https://downloads.rclone.org/v1.47.0/rclone-v1.47.0-linux-amd64.deb"

# copy local files
COPY root/ /root-layer/

## Single layer deployed image ##
FROM scratch

# Add files from buildstage
COPY --from=buildstage /root-layer/ /

Here we are leveraging a multi stage DockerFile to run custom logic and pull down an Rclone deb from the Internet to include in our image layer for distribution. Any amount of logic can be run in this build stage or even multiple build stages as long as the files in the end are combined into a single folder for the COPY command in the final output.

Getting a Mod to Dockerhub

To publish a Mod to DockerHub you will need the following accounts:

We recommend using this repository as a template for your first Mod, so in this section we assume the code is finished and we will only concentrate on plugging into GitHub Actions/Dockerhub.

The only code change you need to make to the build logic file .github/workflows/BuildImage.yml will be to modify the ENDPOINT to your own image:

  ENDPOINT: "user/endpoint"
  BRANCH: "master"

User is your Dockerhub user and endpoint is your own custom name (typically the name of the repository where your mod is). You do not need to create this endpoint beforehand, the build logic will push it and create it on first run.

Head over to https://github.com/user/endpoint/settings/secrets and click on New secret

Add DOCKERUSER (your DockerHub username) and DOCKERPASS (your DockerHub password or token).

You can create a token by visiting https://hub.docker.com/settings/security

GitHub Actions will trigger a build off of your repo when you commit. The image will be pushed to Dockerhub on success. This Dockerhub endpoint is the Mod variable you can use to customize your container now.

Getting a Mod to GitHub Container Registry

To publish a Mod to GitHub Container Registry you will need the following accounts:

We recommend using this repository as a template for your first Mod, so in this section we assume the code is finished and we will only concentrate on plugging into GitHub Actions/GitHub Container Registry.

The only code change you need to make to the build logic file .github/workflows/BuildImage.yml will be to modify the ENDPOINT to your own image:

  ENDPOINT: "user/endpoint"
  BRANCH: "master"

User is your GitHub user and endpoint is your own custom name (typically the name of the repository where your mod is). You do not need to create this endpoint beforehand, the build logic will push it and create it on first run.

Head over to https://github.com/user/endpoint/settings/secrets and click on New secret

Add CR_USER (your GitHub username) and CR_PAT (a personal access token with read:packages and write:packages scopes).

You can create a personal access token by visiting https://github.com/settings/tokens

GitHub Actions will trigger a build off of your repo when you commit. The image will be pushed to GitHub Container Registry on success. This GitHub Container Registry endpoint is the Mod variable you can use to customize your container now.

Submitting a PR for a Mod to be added to the official LinuxServer.io repo

  • Fork this repo, checkout the template branch.
  • Edit the Dockerfile for the mod. Dockerfile.complex is only an example and included for reference; it should be deleted when done.
  • Inspect the root folder contents. Edit, add and remove as necessary.
  • Edit the readme with pertinent info.
  • Finally edit the .github/workflows/BuildImage.yml. Customize the vars for BASEIMAGE and MODNAME.
  • Ask the team to create a new branch named <baseimagename>-<modname> in this repo. Baseimage should be the name of the image the mod will be applied to. The new branch will be based on the template branch.
  • Submit PR against the branch created by the team.
  • Make sure that the commits in the PR are squashed.
  • Also make sure that the commit and PR titles are in the format of <imagename>: <modname> <very brief description like "initial release" or "update">. Detailed description and further info should be provided in the body (ie. code-server: python2 add python-pip).

Appendix

Inspecting mods

To inspect the file contents of external Mods dive is a great CLI tool:

https://github.com/wagoodman/dive

Basic usage:

docker run --rm -it \
    -v /var/run/docker.sock:/var/run/docker.sock \
    wagoodman/dive:latest <Image Name>