Skip to content


Switch branches/tags

Name already in use

A tag already exists with the provided branch name. Many Git commands accept both tag and branch names, so creating this branch may cause unexpected behavior. Are you sure you want to create this branch?

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time


Garrison is a simple tool I use for deploying docker containers to large amounts of servers, mostly on EC2.

The idea behind it is that most people just need a few bash scripts to be run on a lot of servers, and don't want to mess around with larger scale "Cloud" solutions.

If you're using auto-scaling or something like that, then this probably isn't for you. Garrison uses a push model, where you probably need a pull model. Consider just setting up your build scripts as "User Data"?


  • golang
  • ssh


go get


Typically, you would create a garrison.json file in your project root. This describes what your infrastructure looks like. You have configuration for each server, and tasks that can be run on each server.


Tasks can only be scripts at the moment. They can be run in parallel, and can have environment variables.

When a set of parallel tasks are run, you won't see any output until the first tasks complete. Output from tasks will only show when each task completes.

	"name": "your-task-name",
	"script": "/some-local-file",
	"parallel": false,
	"environment": {
		"ENV_VAR1": "value1",
		"ENV_VAR2": "value2"

Task parameters

Sometimes you might need to pass an environment variable through to some task and you don't want to have to change your configuration every time. For example, you may be telling your build script which git branch you want to build.

You can specify parameters in your task like this:

  "name": "build",
  "parameters": ["BRANCH"],
  "script": "build"

And then execute garrison as usual and it will use the BRANCH environment variable:

BRANCH=some-branch garrison group:task

Example with Elasticsearch

Lets say we want to deploy andrewvos/docker-elasticsearch. We need a task to build (docker pull) the elasticsearch image on each server, and we need a task to launch the container on each server.

This is what our garrison.json file looks like:

		"name": "elastic",
		"servers": [
			{ "address": "", "user": "ubuntu" },
			{ "address": "", "user": "ubuntu" }
		"tasks": [
			{ "name": "build", "script": "deployment-scripts/build", "parallel": true },
			{ "name": "launch", "script": "deployment-scripts/launch" }

Notice that parallel is specified for the build task. This is because we want this task to run as fast as possible. Garrison will execute the build script on all servers at the same time. For the launch task we want to be sure that at least one elasticsearch server is up at a time, so we don't execute this in parallel.

And we have a bash script for build:

#!/bin/bash -e
docker pull andrewvos/docker-elasticsearch

And one for launch:

export CONTAINER=elasticsearch
docker stop $CONTAINER || :
docker rm $CONTAINER || :
docker run -d --name $CONTAINER -p 9200:9200 -p 9300:9300 -v /some/path:/var/lib/elasticsearch andrewvos/docker-elasticsearch

To launch the build task just run garrison elastic:build. Note that you can also single out a specific server by using the address or the index in the task name:

garrison elastic:0:build
garrison elastic:

ZSH Completion

Add the following to your .zprofile.

fpath=($GOPATH/src/ $fpath)


A kind of deployment tool thingy






No releases published


No packages published