In a ceph cluster, how do we replace failed disks while keeping the osd id(s)? Here are the steps followed (unsuccessful):
