This guide records the playbook we followed on 2025-11-01 when Ceph started warning that each OSD hosted too many placement groups (PGs). It assumes you have kubectl access to the rook-ceph namespace and that the toolbox pod (deploy/rook-ceph-tools) is available.
KUBECONFIG=~/home-ops/kubeconfig \