NYCPHP Meetup

[Rob Marscher]

>> On 3/6/2011 10:26 PM, Matt Juszczak wrote:
>>> Yes, but most nosql databases aren't designed for durability or reliable persistence.

The MongoDB developers seem to be tackling this problem (and in general, seem to be committed to make their db the best of both worlds).  They recommend always having at least two servers with master-slave replication which does a lot to prevent potential data loss.  But there's also things you can do with mongo to confirm that your data got written disk before the client moves on to the next command.  So if there are operations where speed is the most important factor (like logging) then you don't do this... but if it's more critical data, then it's worth having the client wait a moment to confirm the write was successful or otherwise retry.

http://www.mongodb.org/display/DOCS/Verifying+Propagation+of+Writes+with+getLastError