Bcdatabase

Bcdatabase is a library and utility which provides database configuration parameter management for Ruby on Rails applications. It provides a simple mechanism for separating database configuration attributes from application source code so that there's no temptation to check passwords into the version control system. And it centralizes the parameters for a single server so that they can be easily shared among multiple applications and easily updated by a single administrator.

$ gem install bcdatabase

A bog-standard Rails application's config/database.yml file looks like this:

development:
  adapter: oracle_enhanced
  database: //localhost/XE
  username: cfg_animal
  password: not-important

test:
  adapter: oracle_enhanced
  database: //localhost/XE
  username: cfg_animal_test
  password: who-cares

production:
  adapter: oracle_enhanced
  database: //super/prod
  username: cfg_animal
  password: very-secret

Rails allows this file to contain ERB. bcdatabase uses ERB to replace an entire configuration block. If you wanted to replace, say, just the production block in this example, you would transform it like so:

<%
  require 'bcdatabase'
  bcdb = Bcdatabase.load
%>

development:
  adapter: oracle_enhanced
  database: //localhost/XE
  username: cfg_animal
  password: not-important

test:
  adapter: oracle_enhanced
  database: //localhost/XE
  username: cfg_animal_test
  password: who-cares

<%= bcdb.production :prod, :cfg_animal %>

This means "create a YAML block for the production environment from the configuration entry named cfg_animal in /etc/nubic/db/prod.yml." The method called can be anything:

<%= bcdb.development :local, :cfg_animal %>
<%= bcdb.staging 'stage', 'cfg_animal' %>
<%= bcdb.automated :dev, :cfg_animal_hudson %>

More rarely, you might need to access the actual configuration hash, instead of the YAMLized version. You can access it by invoking Bcdatabase.load as shown earlier, then using the bracket operator to specify the configuration you want:

bcdb[:local, :cfg_animal]

The resulting hash is suitable for passing to ActiveRecord::Base.establish_connection, for instance.

The database configuration properties for all the applications on a server are stored in one or more files under /etc/nubic/db (by default; see "File locations" below). Each one is a standard YAML file, similar to Rails' database.yml but with a few enhancements:

• Each file can have a defaults entry which provides attributes which are shared across all configurations in the file
• Each entry defaults its "username" attribute to the name of the entry (useful for Oracle)
• Each entry defaults its "database" attribute to the name of the entry (useful for PostgreSQL)

Since each file can define a set of default properties which are shared by all the contained configurations, it makes sense to group databases which have some shared configuration elements.

If you have an /etc/nubic/db/stage.yml file that looks like this:

defaults:
  adapter: oracle_enhanced
  database: //mondo/stage
cfg_animal:
  password: secret
personnel:
  username: pers
  password: more-secret

You have defined two configuration entries. :stage, :cfg_animal:

adapter:  oracle_enhanced
username: cfg_animal
password: secret
database: //mondo/stage

and :stage, :personnel:

adapter:  oracle_enhanced
username: pers
password: more-secret
database: //mondo/stage

Bcdatabase supports storing encrypted passwords instead of the plaintext ones shown in the previous example. Encrypted passwords are defined with the key epassword instead of password. The library will decrypt the epassword value and expose it to the calling code (usually Rails) unencrypted under the password key. The bcdatabase command line utility handles encrypting passwords; see the next section.

While the passwords are technically encrypted, the master key must be stored on the same machine so that they can be decrypted on demand. That means this feature only obscures passwords — it will not deter a determined attacker.

The gem includes a command line utility (also called bcdatabase) which assists with creating epassword entries. It has online help; after installing the gem, try bcdatabase help to read it:

$ bcdatabase help
Tasks:
  bcdatabase encrypt [INPUT [OUTPUT]]  # Encrypt every password in a bcdatabase YAML file
  bcdatabase epass [-]                 # Generate epasswords from database passwords
  bcdatabase gen-key [-]               # Generate the bcdatabase shared key
  bcdatabase help [TASK]               # Describe available tasks or one specific task

/etc/nubic/db is the default place the library will look for the central configuration files. It may be overridden with the environment variable BCDATABASE_PATH. For instance, if you wanted to keep these files in your home directory on your development machine — perhaps so that editing them doesn't require elevated privileges — you could add this to ~/.bashrc:

export BCDATABASE_PATH=${HOME}/nubic/db

Similarly, the file containing the encryption password has a sensible default location, but that location can be overridden by setting BCDATABASE_PASS.

Bcdatabase was originally designed for use with ActiveRecord in Rails applications. Since DataMapper's programmatic configuration mechanism (Datamapper.setup) accepts hashes which are very similar to ActiveRecord configuration hashes, Bcdatabase can easily be used with DataMapper as well. Example:

bcdb = Bcdatabase.load(:transforms => [:datamapper]))
DataMapper.setup(:default, bcdb[:stage, :personnel])

The :datamapper transform passed to Bcdatabase.load in this example permits sharing of one set of Bcdatabase configurations between ActiveRecord and DataMapper-based apps. Specifically, for those cases where the ActiveRecord adapter and the DataMapper adapter have different names, it allows you to specify a separate datamapper_adapter in your Bcdatabase configuration. For example, say you had these contents in stage.yml:

defaults:
  adapter: postgresql
  datamapper_adapter: postgres
personnel:
  password: foo

When loaded without the :datamapper transform, the effective database configuration hash for :stage, :personnel would be

adapter: postgresql
datamapper_adapter: postgres # ignored by AR
database: personnel
username: personnel

With the :datamapper transform, the result would be instead:

adapter: postgres
database: personnel
username: personnel

And so your DM and AR apps can live side-by-side and neither needs to embed its own database credentials.

Bcdatabase works on MRI 1.8.7, 1.9.3, 2.0.0, and 2.1.2. It will also work on JRuby (tested on 1.7), provided that jruby-openssl is also installed. It is continuously tested on all three of these platforms.

There is support for having a different adapter defined for JRuby. It's very similar to the DataMapper support described above: if you define a key called jruby_adapter in a configuration, it will be copied to adapter when Bcdatabase is running under JRuby. Unlike the DataMapper support, it is automatic — no need to specify a transform.

bcdatabase was developed at and for the Northwestern University Biomedical Informatics Center.

Copyright (c) 2009 Rhett Sutphin. See LICENSE for details.