We initially wanted to have a single replica-pair and multiple slaves hinged off that pair, but can’t seem to get that to jive with automatic master failover.

Do I understand correctly that you guys ended up running the entire thing on a single pair of servers then, in the end (unlike the illustration)?

Yea I think it would be good to post something about the developer oriented improvements that have been made recently, as I don’t always feel like people see all of the hard work you guys have been putting in.

Perhaps we can talk a bit more about that at OSCON.

We store all of the project information we need in a single project document in mongoDB. Having a document oriented data store allows us to store all of that data in a very usable format, index it however we want, and retrieve it in a single database hit.

This has proved very valuable, and MongoDB’s replication and peformance have both made it incredibly easy to scale sf.net up as traffic increases.