The problem here occurs with a deployment with a low number of OSD's (say 3) - a number of pools in the ceph radosgw pool setup get 0.1% of the storage available, so the pg_num calculation based on the number of OSD's in the cluster ends up == 0. This results in the error seen (math domain error).
I think we should impose a minimum pg_num to ensure that this does not happen.
The problem here occurs with a deployment with a low number of OSD's (say 3) - a number of pools in the ceph radosgw pool setup get 0.1% of the storage available, so the pg_num calculation based on the number of OSD's in the cluster ends up == 0. This results in the error seen (math domain error).
I think we should impose a minimum pg_num to ensure that this does not happen.