Api Server - The oracle of the dyno world

Introduction

The apiserver is the central server and API endpoint for OpenRuko, it is written in Node.js and uses PostgreSQL to persist state. It exposes methods via a restful API to query and manipulate the state of the system, dyno hosts call in periodically to pick up jobs which we try to distribute fairly by holding on to each request for a small amount of time in order to determine the least burden dyno node. The API tries to aims to be compatible with Heroku client tools, and whilst certain features such as addons, ssl certs aren't support this is an ongoing effort.

Requirements

Tested on Linux 3.2 using node 0.8 and postgreSQL 9.1.

On a fresh Ubuntu 12.04 LTS instance:

PostgreSQL 9.1:
https://help.ubuntu.com/community/PostgreSQL (after installing, createdb root is handy)

apt-get install postgresql postgresql-client postgresql-server-dev-9.1
apt-get install postgresql-contrib-9.1 # required for pycrypto hstore etc..
apt-get install uuid-dev # required by node-uuid 
apt-get install python-software-properties

Node.js 0.8.x is not available in available Ubuntu repositories, however Chris Lea provides a PPA that works very well, following the instructions onsite adding the relevant lines to /etc/apt/sources.list

add-apt-repository ppa:chris-lea/node.js
apt-get update
apt-get install nodejs npm

Please share experiences with CentOS, Fedora, OS X, FreeBSD etc...

Installation

git clone https://github.com/openruko/apiserver.git apiserver  
cd apiserver

Install node.js dependencies:

make init

Create an ssh key

ssh-keygen -t rsa

Create database

sudo -u postgres createuser -s -P -U postgres
# enter your login name
# enter the password, for exemple openruko
createdb openruko

Create certs for the rendezvous endpoint (tls):

make certs

Environment Variables

apiserver/bin/apiserver will check for the presence of several environment variables, these must be configured as part of the process start - e.g. configured in supervisord or as part of boot script see ./apiserver/conf.js

APISERVER_KEY - special key for other services to authenticate with API server (you can generate one with uuidgen)
PG_USER - Your login name, unless you set something else.
PG_PASSWORD - Your postgresql password, for exemple openruko
S3_KEY - You need an Amazon S3 account to store repos and slug archive
S3_SECRET - You need an Amazon S3 account to store repos and slug archive
S3_BUCKET - You need an Amazon S3 account to store repos and slug archive

If you don't want to use Amazon S3, you could use a clone like fakes3. Just add the following environment variable:

S3_HOSTNAME - If yout host an S3 clone on your machine use mymachine.me in place of localhost.
S3_PORT - The S3 clone listening port.

Launch

$ cat > .env << EOF
APISERVER_KEY=xxx
PG_PASSWORD=xxx
S3_KEY=xxx
S3_SECRET=xxx
S3_BUCKET=xxx
EOF

foreman start

Create a user

Once apiserver launched, you will need to create your first user.

./apiserver/bin/adduser

Test

createdb openruko_test
PG_PASSWORD=xxx npm test

Help and Todo

Loads to do on this part, not limited to:

Improve setup scripts and find a path to database migrations.
Missing features
In updatestate handle logic to handle failed instance, retries, and kills - if dyno is not consistent with state in database.

License

apiserver and other openruko components are licensed under MIT.
http://opensource.org/licenses/mit-license.php

Authors and Credits

Matt Freeman
email me - im looking for some remote work
follow me on twitter

srobertson/apiserver