dstore

Introduction

dstore is an interface for different storage formats and databases (let's call them datastores from now on).
It takes 1 kind of input, and translates this into the correct format for every storage engine.

Besides abstracting away the input format, dstore also provides a single notation for your data's schema (called blueprints).

Blueprints tell dstore what your data looks like.

This is used for:

• Validation of the input when storing items (via json schema).
  • json schema http://spacetelescope.github.io/understanding-json-schema/
  • built-in schemas https://github.com/trappsnl/dstore/tree/master/schemas
• Serialization of the input when sending the item to a specific store.
• Creation of the PostgreSQL tables / creation of the Elasticsearch mappings.
• Transforming of items to make it compatible with blueprints of different versions.

dstore also tracks the changes you make to your blueprints.
By knowing what has changed, it is able to make the data compatible with older and never versions of your project.
This means old clients will still receive the latest data, and data that gets inserted by older clients is still available to new clients (although maybe lacking some "columns" in the data, in case new columns were added).

Currently, dstore supports PostgreSQL, Elasticsearch and LevelDB, a great stack for a modern web application.

Overview

Topics

• Projects
• Blueprints
• Items
• API
• Requirements
• Installation
• Dive deeper

Projects

Projects are similar to a "database" or an "elasticsearch index", In case of dstore, a project is nothing more than an identifier under which we store blueprints and version information. A good identifier for your project is a reverse-DNS
or a simple string like "blog" or "website"

Blueprints

The blueprint describes your data format, so the stores know what data they can expect and know how to serialize it.
A blueprint contains information like the table name, elasticsearch type, the columns and the validation options that should be used when data is stored.
Let's look at an example how to create a blueprint for storing posts for a blog. For this, we use Put blueprint command, and use "myblog" as the project identifier, and "article" as the blueprint identifier.

curl -X PUT http://localhost:2020/myblog/article/_blueprint -d '
{
  "postgresql": {
    "table": "articles"
  },
  "elasticsearch": {
    "type": "article"
  },
  "columns": {
    "title_nl": {
      "type": "string"
    },
    "title_en": {
      "type": "string"
    },
    "intro_nl": {
      "type": "text"
    },
    "intro_en": {
      "type": "text"
    },
    "content_nl": {
      "type": "text"
    },
    "content_en": {
      "type": "text"
    },
    "date_created": {
      "type": "datetime",
      "validation": {
        "required": true
      }
    },
    "date_changed": {
      "type": "datetime",
      "validation": {
        "required": true
      }
    }
  }
}'

This command created a "myblog" project, containing a "article" blueprint. All we have to do next is tag the "myblog" project to make it ready for accepting data. But first, take a look at the column types that dstore supports and what they translate to for the different datastores.

column type	postgresql type	elasticsearch type	leveldb type
uuid	UUID	string	String (JSON)
uuid[]	UUID[]	string	Array (JSON)
string	STRING	string	String (JSON)
string[]	STRING[]	string	Array (JSON)
text	TEXT	text	String (JSON)
text[]	TEXT[]	text	Array (JSON)
point	GEOMETRY(Point, 4326)	geo_point	Object (GeoJSON)
point[]	GEOMETRY(MultiPoint, 4326)	geo_point	Object (GeoJSON)
linestring	GEOMETRY(LineString, 4326)	geo_shape	Object (GeoJSON)
linestring[]	GEOMETRY(MultiLineString, 4326)	geo_shape	Object (GeoJSON)
polygon	GEOMETRY(Polygon, 4326)	geo_shape	Object (GeoJSON)
polygon[]	GEOMETRY(MultiPolygon, 4326)	geo_shape	Object (GeoJSON)
date	DATE	date (format: yyyy-MM-dd)	String (JSON)
date[]	DATE[]	date (format: yyyy-MM-dd)	Array (JSON)
datetime	TIMESTAMP WITH TIMEZONE	date (format: yyyy-MM-dd HH:mm:ss)	String (JSON)
datetime[]	TIMESTAMP WITH TIMEZONE[]	date (format: yyyy-MM-dd HH:mm:ss)	Array (JSON)
float	REAL	float	Number (JSON)
float[]	REAL[]	float	Array (JSON)
integer	INTEGER	integer	Number (JSON)
integer[]	INTEGER[]	integer	Array (JSON)
boolean	BOOLEAN	boolean	Boolean (JSON)
boolean[]	BOOLEAN[]	boolean	Array (JSON)
json	JSON	object	Object (JSON)
json[]	JSON[]	object	Array (JSON)

As you can see, we follow PostgreSQL's [] notation for defining an array of something.

Versions

When you are done adding blueprints to a project, it's time to create a project tag. By creating a tag we are saving the current state of all blueprints, and assign a tag version number to it. The migrators for every store will kick into action to prepare themselves for accepting data in this new format.
In practice, this means that a new database and elasticsearch index is created with it's name being a combination of your project's identifier, a "v" and the tagged version. (e.g. myblogv1, myblogv2 etc...).
For LevelDB, it's quite easy. Since it's schemaless we don't have to migrate anything.

You can create a tag with the Tag project command:

curl -X POST http://localhost:2020/myblog/_version

When the request completes, the storage engines are ready to handle data with the new blueprint.

Items

Storing items is done via a simple PUT command. The request body is JSON and should, at the very least contain the following keys:

• project_version An existing project tag
• id A UUID that does or does not yet exist in the database

You can also include a links key that is an array of UUID's, pointing to other items
Internally (and above), we refer to data as an "item", this is the same concept as a elasticsearch document or a table row.

Below is an example:

curl -X PUT http://localhost:2020/myblog/article/66276124-ebcd-45e1-8013-825346daa283 -d '
{
  "id": "66276124-ebcd-45e1-8013-825346daa283",
  "project_version": 1,
  "title_nl": "De titel",
  "title_en": "Some title",
  "intro_nl": "De intro",
  "intro_en": "The intro",
  "content_nl": "De inhoud",
  "content_en": "The content",
  "date_created": "2014-01-17 03:50:12",
  "date_changed": "2014-01-17 03:50:12"
}'

Deleting an item is not so difficult either:

curl -X DELETE http://localhost:2020/myblog/article/66276124-ebcd-45e1-8013-825346daa283

API

At this moment, the only way to communicate with dstore is via a JSON API.
We are looking to add more ways of communicating with dstore in the future.

Check the API documentation over at apiary.io for a full overview of possibilities to manage and retrieve information about projects, blueprints and version.

Requirements

• Node.js
• RabbitMQ
• PostgreSQL
• PostgreSQL contrib modules (for UUID support)
• Postgis (for spatial support)
• A "template_postgis" template with uuid and/or postgis support
• A PostgreSQL user with premissions for creating databases
• Elasticsearch

To run the script, you must make the following environment variables available.

export POSTGRESQL_HOST="localhost"
export POSTGRESQL_PORT="5432"
export POSTGRESQL_USER="..."
export POSTGRESQL_PASSWORD="..."
export ELASTICSEARCH_HOST="http://localhost:9200"
export QUEUE_CONNECTIONSTRING="amqp://guest:guest@localhost:5672"
export PROJECTOR_PATH=`pwd`
export LEVEL_PATH="$PROJECTOR_PATH/storage/level"
export PROJECT_FILE_PATH="$PROJECTOR_PATH/storage/blueprint"
export PORT=2020

Installation

We build a .deb file that installs dstore on your system. It is made & tested on Ubuntu 14.04, but probably works in debian as well. It will go through all the instructions as seen in DIY.

Vagrant

git clone https://github.com/trappsnl/dstore.git
cd dstore
vagrant up

APT

wget https://github.com/trappsnl/dstore/raw/master/build/debinstall/dstore-1.deb
sudo dpkg -i dstore-1.deb

# missing dependencies ?
# if you don't already have elasticsearch installed, add the repository as described below
wget -qO - https://packages.elasticsearch.org/GPG-KEY-elasticsearch | sudo apt-key add -
sudo add-apt-repository "deb http://packages.elasticsearch.org/elasticsearch/1.4/debian stable main"
sudo apt-get update

# you can now install all missing dependencies them like this:
sudo apt-get -f install

# now try again
sudo dpkg -i dstore-1.deb

DIY

# install dstore
npm install --save dstore

# install PM2 (node.js process manager)
sudo npm install -g pm2

# add elasticsearch repository
wget -qO - https://packages.elasticsearch.org/GPG-KEY-elasticsearch | sudo apt-key add -
sudo add-apt-repository "deb http://packages.elasticsearch.org/elasticsearch/1.4/debian stable main"
sudo apt-get update

# install dependencies
sudo apt-get install rabbitmq-server postgresql-9.3 postgresql-contrib postgresql-9.3-postgis-2.1 nodejs build-essential openjdk-7-jdk libpq-dev

# change user to postgres
sudo su postgres

# create postgresql database
createdb -E UTF8 -T template0 template_postgis

# create postgis template for postgresql (only if you need spatial support)
psql template_postgis <<EOF
CREATE EXTENSION "uuid-ossp";
CREATE EXTENSION postgis;
UPDATE pg_database SET datistemplate = TRUE WHERE datname = 'template_postgis';
EOF

# export necessary config variables
export POSTGRESQL_HOST="localhost"
export POSTGRESQL_PORT="5432"
export POSTGRESQL_USER="..."
export POSTGRESQL_PASSWORD="..."
export ELASTICSEARCH_HOST="http://localhost:9200"
export QUEUE_CONNECTIONSTRING="amqp://guest:guest@localhost:5672"
export PROJECTOR_PATH=`pwd`
export LEVEL_PATH="$PROJECTOR_PATH/storage/level"
export PROJECT_FILE_PATH="$PROJECTOR_PATH/storage/blueprint"
export PORT=2020

# start dstore
cd path/to/dstore/bin && ./start.sh

#Dive deeper Head over to the Api docs to learn more about the internals.