We utilize Ceph for most of the storage-needs in our openstack deployment. Ceph is a properly scalable and resillient storage-solution which run on regular servers. The following articles describes various bits on how me manage our Ceph solutions:

Maintainance

Ceph is a «automatically healing» storage-cluster, which makes sure to rebalance data if any disks or storage-nodes are unavailable. During maintainance-windows it might be desirable to disable this functionality. This is done by instructing ceph to not consider unavailable disks being out of the cluster with the 'noout' flag. The flag can be set/unset globally on the cluster:

$ sudo ceph osd set noout
$ sudo ceph osd unset noout

Or it can be set for OSDs on a certain host:

$ sudo ceph osd add-noout <storage-node>
$ sudo ceph osd rm-noout <storage-node>