Infrastructure » website

There is a big mess with pg count for pool creation in different places in docs, moreover, some of them cause ceph pool create to complain.

Page1: https://docs.ceph.com/docs/jewel/rados/operations/placement-groups/#set-the-number-of-placement-groups
Less than 5 OSDs set pg_num to 128
Between 5 and 10 OSDs set pg_num to 512
Between 10 and 50 OSDs set pg_num to 4096

Page2: https://docs.ceph.com/docs/jewel/rados/operations/pools/#create-a-pool
osd pool default pg num = 100
osd pool default pgp num = 100
(which is not power of two!)

cluster with 12 OSD is >10, so it should be 4096, but ceph rejects it:
ceph --cluster ceph osd pool create test_pool 4096 4096
Error ERANGE: pg_num 4096 size 3 would mean 12291 total pgs, which exceeds max 3000 (mon_max_pg_per_osd 250 * num_in_osds 12)

There is a strong requirements to have power of two of PGs, but default mon_max_pg_per_osd is set to 250 (which is not power of two).

Page3: https://docs.ceph.com/docs/jewel/rados/operations/pools/
It says 'When you create a pool, set the number of placement groups to a reasonable value (e.g., 100).' (again, not a power of two).