Previously the helm overrides for nova, cinder and glance respectively
defined 512, 256 and 256 as the PG_NUM for their Ceph pools. Until
Mimic, ceph would just issue a warning message if this number was
bigger than the number OSD * 100, now Nautilus returns an error
message
and asks for expected_num_objects parameter or --yes-i-really-mean-it.
Neither of these options is supported by openstack-helm.
The pool creation is adjusted to take into account the number of OSDs
available to choose the PG_NUM.
The steps are:
Number of OSD times 100 minus 1;
Get the nearest power of two numbers below this result;
Limit PG_NUM to the previous defaults, as they are already high
numbers.
This logic roughly results in the same values described here: https://docs.ceph.com/en/nautilus/rados/operations/placement-groups/#a-preselection-of-pg-num
This was done to solve an error message which demands a
expected_num_objects when PG_NUM is considered too high by ceph.
Test plan:
StarlingX builds successfully
stx-openstack is built successfully
stx-openstack is applied successfully and ceph pools are created
accordingly
It is possible to override PG_NUM on ceph pools by changing chunk_size
value through helm overrides
Reviewed: https:/ /review. opendev. org/c/starlingx /config/ +/817776 /opendev. org/starlingx/ config/ commit/ c861bc6f34ee722 462ba98bbde8808 d23e602733
Committed: https:/
Submitter: "Zuul (22348)"
Branch: master
commit c861bc6f34ee722 462ba98bbde8808 d23e602733
Author: Delfino Curado <email address hidden>
Date: Fri Nov 12 10:21:43 2021 -0500
Fix ceph pool creation during stx-openstack apply
Previously the helm overrides for nova, cinder and glance respectively num_objects parameter or --yes-i- really- mean-it.
defined 512, 256 and 256 as the PG_NUM for their Ceph pools. Until
Mimic, ceph would just issue a warning message if this number was
bigger than the number OSD * 100, now Nautilus returns an error
message
and asks for expected_
Neither of these options is supported by openstack-helm.
The pool creation is adjusted to take into account the number of OSDs /docs.ceph. com/en/ nautilus/ rados/operation s/placement- groups/ #a-preselection -of-pg- num
available to choose the PG_NUM.
The steps are:
Number of OSD times 100 minus 1;
Get the nearest power of two numbers below this result;
Limit PG_NUM to the previous defaults, as they are already high
numbers.
This logic roughly results in the same values described here:
https:/
This was done to solve an error message which demands a num_objects when PG_NUM is considered too high by ceph.
expected_
Test plan:
StarlingX builds successfully
stx-openstack is built successfully
stx-openstack is applied successfully and ceph pools are created
accordingly
It is possible to override PG_NUM on ceph pools by changing chunk_size
value through helm overrides
Closes-Bug: #1949360 a95a3706c72dd21 b8ed3efbe60
Signed-off-by: Delfino Curado <email address hidden>
Change-Id: I222bee29bcaa09