Ceph

To minimize the recovery time of the cluster in the event of an OSD failure it would be better to recover to the remaining OSD's instead of recovering to the replaced OSD when the replaced OSD comes with no data (disk crash).
Gregory came with an elegant solution to set the weight of the replaced OSD to a very low number. It could be set to 0 for the time the cluster recovers and increased gradually after the cluster recovered to prevent the OSD of overloading.
For full information see the IRC discussion:
http://irclogs.ceph.widodh.nl/index.php?date=2012-11-06 from [21:12]