Hi all, I think of releasing Sheepdog 0.2.0 at the end of this month, Dec 31st 2010. In this version, I think of solving critical issues for reliability. The features which are not in the tree yet are: - recovery from total node failure - update metadata atomically (journaling) - some error handlings (e.g. connection timeout, disk full, ...) - fix races when accessing objects during a node membership change If possible, I'd like to add a multi-ring membership patch to solve scalability problem, too. If you have pending patches, please send until Dec 28th 2010. Thanks, Kazutaka |