Ceph » RADOS

Consider this output:
GLOBAL:
SIZE AVAIL RAW USED %RAW USED
30.2G 14.3G 15.9G 52.78
POOLS:
NAME ID USED %USED MAX AVAIL OBJECTS
cephfs_data_a 1 0 0 4.65G 0
cephfs_metadata_a 2 2.19K 0 4.65G 21
foo 3 4.25G 23.33 4.65G 1088

The cluster is ~50% full but pool foo only uses ~23% of its max capacity. All pools have 3 replicas.
To reproduce:

../src/vstart.sh -n -s -d
bin/ceph osd pool create foo 8 8
bin/rados bench -p foo 60 write --max-object-size 4K --no-cleanup
bin/ceph df