Any plans to provide support for rolling upgrade? Upgrading clusters with zero-downtime is necessary for us.