In 64-bit mongodb how AutomaticFailover works,when each shard has different data?

I am handling with Autosharding,and i had questioned about Whether the data in the shard "A" Will be available in shard "B".They have answered as Data in the shard "A" Will not be available in shard "B".In this scenario,how the automatic failover works? For example i have 3 Shards one of my shards gets failed, Then we can access data from the other shards right?. If the data is different in each shard then how can we a开发者_JAVA百科ccess data?...Anyone can explain about this..Plz..

Sharding is not about failover but rather about scalability. Failover achieved with replica sets. I.e. each shard is running as replica set with multiple nodes, when the master node fails new master node is elected among slave ones.

Here how it looks: http://www.infoq.com/resource/news/2010/08/MongoDB-1.6/en/resources/mongodb2.png