Docker Git Deploy

Dockerized Git deploy through SSH service, built on top of official Ubuntu trusty image. It uses git 'post-receive' hook and provides nice colored logs for each pushed commit.

Accepts an IN, BRANCH_*, USER and PUBLIC_KEY settings through environment variables, if the git history doesn't matter to you, pass only the BRANCH_* and PUBLIC_KEY settings, which aren't optional

This Dockerized image doesn't allow plain text logins, can only connect to it through the use of a public RSA key.

Defaults

BRANCH_* = "" # [REQUIRED] The folder that receives the git checkout depending on the name of the variable
# use BRANCH_MASTER="/out" for example, or BRANCH_TESTING="/testout"

PUBLIC_KEY = "" # [REQUIRED] Your mounted public key path inside the container

USER = git # [OPTIONAL] The user used in the git push

IN = "" # [OPTIONAL] The folder that holds the git bare repo

Setup

$ docker run -d \
    -p 1234:22 \
    --name deploy \
    -e PUBLIC_KEY="/id_rsa.pub" \
    -v ~/.ssh/id_rsa.pub:/id_rsa.pub \
    -e BRANCH_TRUSTY="/out" \
    -e BRANCH_MASTER="/out" \
    -v /var/www:/out \
    -e IN="/in" \
    pocesar/docker-git-deploy

c48f7b86594953012ca4731b1ec08b053ce5826d3f501ed579c660bec42d2c88

NOTE: You can use -e PUBLIC_KEY="$(cat ~/.ssh/id_rsa.pub)" as well and drop the -v part of the public key

Check the examples folder with a sample docker-compose.yml with a Typescript / Node.js project

Deploy

git remote add upstream ssh://docker@yourhost:1234/in # or /home/git/repo.git by default
git commit -m "Behold!"
git push upstream master

Get logs (with colors!)

$ docker logs deploy
[+] 2014-09-18T11:36:05Z: Using branches:
         - BRANCH_TRUSTY=/out
         - BRANCH_MASTER=/out

[+] 2014-09-18T11:36:05Z: Reading public key mount
[+] 2014-09-18T11:36:05Z: Created user docker
[+] 2014-09-18T11:36:05Z: Using existing path
Initialized empty shared Git repository in /in/
[+] 2014-09-18T11:36:05Z: Deploy using this git remote url: ssh://docker@host:port/in
Server listening on 0.0.0.0 port 22.
Server listening on :: port 22.
Accepted publickey for docker from 172.17.42.1 port 35341 ssh2: RSA 79:4f:46:33:1f:39:25:6d:0d:37:e1:e0:d2:42:5c:0e
[^] 2014-09-18T11:36:05Z: Updated sources on BRANCH_TRUSTY:/out
-------------
trusty a4f9d05 - Paulo Cesar, 27 hours ago: README
-------------
Received disconnect from 172.17.42.1: 11: disconnected by user
Accepted publickey for docker from 172.17.42.1 port 35344 ssh2: RSA 79:4f:46:33:1f:39:25:6d:0d:37:e1:e0:d2:42:5c:0e
[^] 2014-09-18T11:36:05Z: Updated sources on BRANCH_MASTER:/out
-------------
master a4f9d05 - Paulo Cesar, 27 hours ago: README
-------------
Received disconnect from 172.17.42.1: 11: disconnected by user

Mods

You can inject your bash scripts into the post-receive hook by mounting your script to -v /var/some/script.sh:/userscript. It can be anything, like set folder / file permissions, execute npm install, bower install etc. and any type of language (if you install the needed binaries of course).

If you mount a script to -v /var/www/somescript.sh:/failscript, in case your /userscript fails, this will be called. this is useful to do some cleanup or revert a commit in case of failure. a simple revert command is git reset --hard HEAD~1. /failscript receive the same arguments that /userscript does

You can also use echo "hello world" >> $MEM_LOG to output stuff to the docker log from any of your scripts. Be aware that the /userscript will be called everytime there's a push.

The setup script, placed in /setup, set as -v /var/some/script.sh:/setup, it will be called one time when the container starts. Useful to install extra software you may need (like ruby, node, jekyll). Only the DOCKER set environment variables are available inside (like BRANCH_*, MEM_LOG, IN, USER, GIT_DIR, HOME). Eg.:

#!/bin/bash

# this is /setup

apt-get -y -qq install nodejs
wget http://example.com/something.js > "$BRANCH_MASTER/something.js"
echo "something.js downloaded" >> $MEM_LOG # goes to docker logs

#!/usr/bin/env node

/** this is /userscript
 * it receives the following arguments:
 * $1 = the branch name (as seen in git)
 * $2 = the refname, like ea9da19
 * $3 = the current path of the working dir
 *
 * all environment variables are available inside, including the IN, BRANCH_* and GIT
 */

var
    path = require('path'), fs = require('fs');

# process.argv[1] === branch
# process.argv[2] === gitref
# process.argv[3] === working tree

if (process.argv[1] === 'master') {
    fs.chmodSync(path.join(process.env.BRANCH_MASTER, 'cache'), '0774');
}