READ_PREFERENCE.md

Read Preference in Ruby

Setting the read preference

You can using the :read option to specify a query's read preference. There are for now two possible options:

@collection.find({:doc => 'foo'}, :read => :primary)
@collection.find({:doc => 'foo'}, :read => :secondary)

In the first case, the query will be directed to the primary node in a replica set. In the second, the query will be sent to a secondary node. The driver will attempt to choose a secondary node that's nearby, as determined by ping time. If more than one secondary node is closeby (e.g, responds to pings within 10ms), then a random node within this subset will be chosen.

Read preference inheritance

The Ruby driver allows you to set read preference on each of four levels: the connection, database, collection, and cursor (or read operation). Objects will inherit the default read preference from their parents. Thus, if you set a read preference of {:read => :secondary} when creating a new connection, then all databases and collections created from that connection will inherit the same setting. See this code example:

@con = Mongo::ReplSetConnection.new(['localhost:27017','localhost:27018'], :read => :secondary)
@db  = @con['test']
@collection = @db['foo']
@collection.find({:name => 'foo'})

@collection.find({:name => 'bar'}, :read => :primary)

Here, the first call to Collection#find will use the inherited read preference, {:read => :secondary}. But the second call to Collection#find overrides this setting by setting the preference to :primary.

You can examine the read preference on any object by calling its read_preference method:

@con.read_preference
@db.read_preference
@collection.read_preference

Future work

In the v2.0 release of the driver, you'll also be able to specify a read preference consisting of a set of tags. This way, you'll be able to direct reads to a replica set member. You can follow this issue's progress here: (https://jira.mongodb.org/browse/RUBY-326).