https://github.com/rook/rook/issues/6948

NS=rook-ceph
CEPHCMD="kubectl -n $NS exec rook-ceph-toolbox-XXX -- "
$CEPHCMD  osd out $OSD_ID
kubectl -n $NS delete pvc $PVC_NAME --wait=false
kubectl -n $NS delete deployment $OSD_DEPLOYMENT
kubectl -n $NS delete secrets $OSD_AUTH
kubectl -n $NS delete job $OSD_PREPARE_JOB
$CEPHCMD auth del osd.$OSD_ID
$CEPHCMD osd purge $OSD_ID --yes-i-really-mean-it
kubectl -n $NS delete pod $(kubectl -n $NS get pod -lapp=rook-ceph-operator -ojsonpath={'.items[0].metadata.name}')
I consider at least one of the following two changes is necessary.

Describe the above-mentioned steps in the ceph-osd-mgmt.md.
Provide the way to remove an arbitrary OSD removal.
Eventually, it would be better to do both changes.