jgwehr / homelab-docker

Docker Compose for building a home lab
106 stars 15 forks source link
caddy crowdsec docker-compose jellyfin self-hosted

A personal project to provide security, privacy, and data-ownership for my home.

GitHub last commit GitHub Repo Stars

Includes:

Homepage

Technology

The following apps / technologies are grouped into ./services/.

Service Note Includes
adblock-and-dns Pihole Pihole Unbound Unbound
change-detect ChangeDetection.io ChangeDetection.io Chrome Chrome
dashboard Homepage Homepage
downloads qBitTorrent qBitTorrent Gluetun Gluetun Radarr Radarr Sonarr Sonarr Lidarr Lidarr Prowlaar Prowlaar + PodGrab
events Rallly Rallly PostGres PostGres
gloomhaven Gloomhaven SecretariatGloomhaven Secretariat
infra Required docker socket proxy
image-board Pinry Pinry
media-request Jellyseerr Jellyseerr
media-streaming Jellyfin Jellyfin
monitor Dozzle Dozzle Diun Diun Scrutiny Scrutiny Speedtest Tracker Speedtest Tracker Uptime Kuma Uptime Kuma
paperless Paperless ngx Paperless-ngx PostGres PostGres
public Reverse Proxy and DDNS Caddy Caddy DuckDNS DuckDNS
recipes Tandoor Recipes Tandoor Recipes PostGres PostGres
security Run this if you're running public CrowdSec CrowdSec + Endlessh

Explanation

Groupings are chosen based on context and dependencies. Most Services "stacks" should work fine without any other services (some exceptions exist) - allowing you to take from this repo what you care about and ignore the rest.

  1. All Services use the same .env file from the root directory (/opt/docker/homelab-docker/.env). This simplifies maintenance as some variables are used by multiple Services. This is accomplished via symlink.
  2. One-way dependency: Stack A can be dependent on Container(s) within Stack B, but Stack B cannot also be dependent on any Containers in Stack A. Similarly, Container A.1 can depend on Container A.2 but Container A.2 cannot also be dependent on Container A.1.
  3. Multiple dependencies: Containers can have any number of (efficient) dependencies. Container A.1 can depend on Container A.2 and Container B.1. Care should be taken to avoid circular dependencies (A on B on C on A).
  4. Services are easy to manage (start, stop, restart,...) independently from one another. Additionally, some apps (eg. Dozzle) make good use of separate docker-compose files to improve their own functionality.

Starting services

  1. Ensure you've established the symlink to .env. See or execute ./scripts/run-each-update.sh
  2. Navigate to the desired service, eg cd ./services/infra
  3. Execute a standard docker compose up, eg docker compose up -d

Notes

Watchtower is intentionally avoided based off advice from the Selfhosted.show podcast. The idea is to have full control over the versions of containers (rather than automated updates) to improve reliability. Instead, I use Diun and dockcheck.sh currently.

Install / Beginning

WIP

  1. Become familiar with the Project Structure below. This is your "map" of where to expect different things to be. In turn, this will help you configure your setup correctly
  2. SAMBA: After copying smb.conf, customize it as needed: interfaces, share names, users, passwords, etc.

Project Structure

Recommendations via multiple docker files, Where to Put Docker Compose, TRaSH Guides

File System

├── /opt/
│  └── docker/
│     └── homelab-docker/ (this repo)
│        ├── configtemplates/ (for help with non-docker tools eg. Samba or SnapRaid)
│        ├── scripts/ (for help with managing install, backups, etc)
│        ├── services/
│        |  ├── service1/
|        |  |    ├── docker-compose.yml
|        |  |    ├── ~.env (symlink)
|        |  |    ├── dockerfiles/ (for custom builds)
│        |  |    |   └── *.dockerfile (for adhoc builds)
│        |  |    ├── configtemplates/ (service-specific configuration to be copied to config dir, then customized)
│        |  |    └── staticconfig/ (service-specific configuratio. does not move)
|        |  |        ├── container1/
|        |  |        |  └── container-specific-file.ext
|        |  |        └── container*/
│        |  └── service*/
│        └── .env (the master .env file. Each Service symlinks to this)
├── /srv/
│  ├── docker/ (for container's configurations)
│  ├── cache/
│  └── logs/
└── /mnt/storage/
   ├── db/
   ├── staticfiles/
   │  ├── icons/
   │  ├── paperless/
   │  ├── tandoor_media/
   │  └── wallpaper/
   ├── downloads/
   │  ├── audiobooks/
   │  ├── movies/
   │  ├── music/
   │  ├── paperless/ (SAMBA shared r/w for ingestion)
   │  ├── podcasts/
   │  └── tv/
   └── media/
      ├── audiobooks/
      ├── music/
      ├── pictures/
      ├── podcasts/
      ├── movies/
      └── tv/

Directories may be created with the following cmds

Please review this script before running it. It is a work in progress and may not run as expected. cd scripts chmod +x start.sh
./start.sh

Establish Directory and File permissions

TBD - this may vary based on your existing file system, user provisions, and/or usage of mergerfs or similar tools. Recursively own the /mnt/storage directory chown -R $USER:docker /mnt/storage (the docker group is available after Docker installation steps) chmod -R a=,a+rX,u+w,g+w /mnt/storage

Docker Compose (and needed files)

/opt explanation: reserved for the installation of add-on application software packages

This github repo represents this folder. It's safely committed to public repos and shouldn't contain anything sensitive.

Example files:

Persistent Data and Configuration

/srv

How you configure the apps and their current states. This is separated from the Docker Compose (ie. "setup") as these become specific to how you use the services - not how they're installed/maintained.

Media Staging (such as magnets)

/mnt/storage/downloads/movies

Media Storage (such as Podcasts, Movies, TV Shows, Audiobooks):

/mnt/storage/media/movies

This structure helps achieve Least Privilege by separating concerns as efficiently as possible. Clear organization can also help with backup prioritization.

Examples:

Nesting the media adjacent to downloads is suggested via servarr.com as a way to allow atomic file moves as opposed to a more intensive/longer copy+paste action. Explained here. Importantly, I configure these services to hardlink finished downloads - this preserves seeding ability while having no effect on storage consumption.

Container Roles and Access to Files

Let's recognize four kinds of Media Server roles containers/apps:

  1. A Curator whose role is to enable the discovery and selection of Media by an End User. It then delivers a Work Order to the Acquirer. It also monitors the Media and ensures it is accurately described and and of desired quality. Ex. Sonarr, Radarr, Lidarr, Readarr, Ombi
  2. An Indexer whose role is to interpret how a Work Order should be completed. It figures out where the Media should come from while negotiating with each supplier. Ex. Jackett, Prowlaar
  3. An Acquirer whose role is to accept and execute a Work Order to retrieve a specific Media. It then delivers that Media to the Provider. Ex. qBitTorrent, Transmission
  4. A Provider whose role is to provide Media (movies, podcasts, etc) to end users. It is unconcerned with how the Media came to exist and isn't responsible for its quality or description. Ex. Emby, Plex, Jellyfin

Port Reservations

Ports are controlled through variables (ie. .env) to provide a central "fact check".

Optionals

Caddy and DuckDns.org