I am in the middle of evaluating HyperDex. It's still early stages, but it's very promising when you want to slice and dice your data on multiple dimensions.
Having said that, I had trouble stopping and/or restarting the cluster in a clean way. To make me even consider using it on production, it should also offer me ways to backup data and as well as convince me that future upgrades will be somewhat smooth.
Not sure when you last looked at it, but we added the ability to cleanly shutdown and restart the cluster two releases ago. We are committed to providing a smooth upgrade path as well; that is not to say we will always be binary-compatible, as the next release involving the disk layer will change the on-disk representation, but we will always provide automatic upgrade scripts. So, if you last tested 0.29b or so, do check out the latest code in the repo.
Having said that, I had trouble stopping and/or restarting the cluster in a clean way. To make me even consider using it on production, it should also offer me ways to backup data and as well as convince me that future upgrades will be somewhat smooth.