Ceph Maintenance

Auto-Recovery

Ceph's natural behaviour is to attempt to ensure redundancy of data across available nodes in the even of OSD or node loss.

Maintenance Options

Ceph can be put into a maintenance mode to not follow this behaviour if a node in the cluster needs some maintenance.

Flag Description
noout Prevents OSDs being marked as out of the cluster when they go down, which stops automatic rebalancing
nobackfill Pause backfill operations
norecover Pause all data recovery operations

Apply these flags with the ceph osd set command

ceph osd set noout
ceph osd set nobackfill
ceph osd set norecover

Redundancy

It is important to take into consideration the redundancy of the cluster. Multiple nodes in a cluster may not be taken into maintenance at one time if the resiliency is only configured on a node level.

OSDs going into readonly?

Unset maintenance flags