To understand the need for these commands better, let's assume a scenario where you want to add a new node to your production Ceph cluster. One way is to simply add the new node with several disks to the Ceph cluster, and the cluster will start backfilling and shuffling the data on to the new node. This is fine for a test cluster.

However, the situation becomes very critical when it comes to a production setup, where you should use some of the ceph osd subcommands/flags, which are mentioned as follows, before adding a new node to the cluster, such as noin, nobackfill, and so on. This is done so that your cluster does not immediately start the backfilling process when the new node comes in. You can then unset these flags during non-peak hours, and the cluster will take its time to rebalance:

1. The usages of these flags are as simple as set and unset. For example, to set a flag, use the following command lines:

# ceph osd set <flag_name>
# ceph osd set noout
# ceph osd set nodown
# ceph osd set norecover

2. Now to unset the same flags, use the following command lines:

# ceph osd unset <flag_name>
# ceph osd unset noout
# ceph osd unset nodown
# ceph osd unset norecover