docker-service-manager

Docker Service Manager

This tool allows the user to manage services run within docker containers. I created this tool out of the necessity to have a multitude of services running on a single private server. These services might at times depend on each other and each expose different ports.

After messing around with virtual machines for quite some time, I finally settled on running each service in its own docker container. This application helps me (and maybe you), to easily mount folders and config files, to expose necessity ports and to build and start all necessity images and containers.

It also takes care of setting the containers timezone, so no time mismatches happen between the service and the host.

Structure of a Service

Each service lives inside its own folder. This folder may be located anywhere, and will be symlinked into /etc/docker-services by using

$ docker-service add /some/services/folder

This folder will not be deleted upon uninstalling this module, so you should take care of this on your own!

service.json

A service contains at its heart a service.json file, which may look like this:

{
	"tag": "paulavery/ympd",
	"name": "ympd",
	"ports": [
		8000
	],
	"mounts": {
		"music": "/etc/mopidy/music"
	},
	"deps": [
		"mpd"
	]
}

All its attributes are optional with the exception of a tag which we require.

tag

If a local Dockerfile is found, a new image will be built from it, under this tag. Otherwise, the tag will be pulled from the docker registry.

name

A name to later be used as the services container name. You will use this to reference services (start or stop them as well as have other services depend on them).

If omitted it will default to the part behind the / of your tag. I would suggest to use this property to allow for drop-in replacements of dependencies (e.g. a mopidy and an mpd service could both have different tags but the name mpd, as they are providing the same service for dependents).

ports

An array defining the ports this service will expose to you. This should not contain any ports not exposed to the user.

mounts

Define names for any directories which need to be mounted into your services container. See about the mounts directory below.

deps

An array listing all dependencies of this service. These dependencies will be started before loading up your service.

mounts directory

You should symlink your mounts here under the name defined in your service.json. This allows you to check in the required mounts to git, while easily assigning them on each machine. If you have no mounts, you do not need this directory.

config directory

This directory contains any configuration files for your service. These are linked into the services container, so you may change them without much fuss. You should treat this directory like a Linux systems root folder. So config/etc/something.conf will be linked to /etc/something.conf inside the container. If you have no configuration files, you may omit this folder.

In addition, you may specify entire folders in your service.json's configs property. The following would mount config/home/sabnzbd/.sabnzbd and config/home/sabnzbd/downloads to /home/sabnzbd/.sabnzbd and /home/sabnzbd/downloads respectively:

{
	configs: {
		'home/sabnzbd': {
			'.sabnzbd': {},
			'downloads': {}
		}
	}
}

Single files will still be mounted separately unless they are positioned in a folder which will be mounted.

CLI

The application exposes the following commands:

docker-service start <name>

Builds any necessary images and containers as well as starting all dependencies before starting up your service. First time doing this might take some time.

A -d/--nodaemon flag may be passed, to start a service in non-daemon mode.

docker-service stop <name>

Stops the container running your service

docker-service restart <name>

Restarts a service.

docker-service status <name>

Prints status information as JSON to the command line.

docker-service list

Lists all installed services with their current status.

Errors

This application does NOT print user-friendly error messages. If anything goes wrong (e.g. a dependency is missing), it just throws an error and crashes. The error messages should be self-explanatory though.