Has anybody used this in a production environment? What kind of lessons did you learn? We've been interfacing hadoop and mongodb manually with shell scripts, and we're considering this as a possible solution.
We use the 1.0-SNAPSHOT from January 10 in production without any issues (we had some in the beginning but patches got added), but only in read mode. I haven't tried to write to MongoDB with the connector yet.
We read from a sharded collection on multiple slaves but it's low volume read for now (< 100GB)
We (10gen - I'm the primary author of this integration layer) have had a few people help us do testing with this in production, I don't know who I can name by name at the moment. However, we are happy to help you with any issues you run into and help you integrate cleaner into your stack.
[I can be contacted at my HN name on twitter or brendan at 10gen dot com]