Pool only using one OSD per host

[NullOranje]

I'm trying to setup a simple Ceph cluster using Rook . There are three hosts, with four disks per host. I create a CephBlockPool and storage class.

apiVersion: ceph.rook.io/v1
kind: CephBlockPool
  metadata:
    name: replicapool
    namespace: rook-ceph
  spec:
    replicated:
      size: 3

Everything seems to come up okay, and I'm seeing data replicated across hosts. However, it appears that it only one OSD per host is being used; if I fill that OSD, it does not utilize any others.

I'm new to Rook/Ceph, so I would expect to see it using all the OSDs per host to expand the storage. Where am I going wrong?

[travisn]

The placement on the OSDs is based on Ceph's Placement Groups. They are expected to automatically scale in Rook so that multiple OSDs should be used. Sometimes by default only 1 PG is created, which could cause this issue, and it's not clear why the autoscaler doesn't scale up to more PGs. What do these commands show?

ceph osd pool ls detail
ceph osd df