The autoscaler [1] will be supported during RHCSv4 (though not on by default). This is a request for ceph-ansible to support deployments with this feature this as PG management causes field escalations.
One way this could work is if there's a data structure passed like this:
openstack_pools:
- {"name": backups, "target_size_ratio": 0.1, "pg_autoscale_mode": True, "application": rbd}
- {"name": volumes, "target_size_ratio": 0.5, "pg_autoscale_mode": True, "application": rbd}
- {"name": vms, "target_size_ratio": 0.2, "pg_autoscale_mode": True, "application": rbd}
- {"name": images, "target_size_ratio": 0.2, "pg_autoscale_mode": True, "application": rbd}
Then ceph-ansible could execute commands like this for each pool (e.g. this is for the volumes pool).
$ ceph osd pool create volumes 16
$ rbd pool init volumes
$ ceph osd pool set volumes target_size_ratio .5
$ ceph osd pool set volumes pg_autoscale_mode on
The target_size_ratio is like the percentages customers input into pgcalc. Putting it directly into THT and skipping pgcalc is a better UX.
The 16 in the first command was recommended as a hard coded value above while the name and target_size_ratio would be variables. The reason for 16 was based on Sage's blog [1] where he set 1, but 16 is the usual default and should work with replicated pools while not triggering the overdose protection.
[1] https://ceph.io/rados/new-in-nautilus-pg-merging-and-autotuning/
Comment 1RHEL Program Management
2019-12-11 13:57:40 UTC
Since the problem described in this bug report should be
resolved in a recent advisory, it has been closed with a
resolution of ERRATA.
For information on the advisory, and where to find the updated
files, follow the link below.
If the solution does not work for you, open a new bug report.
https://access.redhat.com/errata/RHSA-2020:2231
The autoscaler [1] will be supported during RHCSv4 (though not on by default). This is a request for ceph-ansible to support deployments with this feature this as PG management causes field escalations. One way this could work is if there's a data structure passed like this: openstack_pools: - {"name": backups, "target_size_ratio": 0.1, "pg_autoscale_mode": True, "application": rbd} - {"name": volumes, "target_size_ratio": 0.5, "pg_autoscale_mode": True, "application": rbd} - {"name": vms, "target_size_ratio": 0.2, "pg_autoscale_mode": True, "application": rbd} - {"name": images, "target_size_ratio": 0.2, "pg_autoscale_mode": True, "application": rbd} Then ceph-ansible could execute commands like this for each pool (e.g. this is for the volumes pool). $ ceph osd pool create volumes 16 $ rbd pool init volumes $ ceph osd pool set volumes target_size_ratio .5 $ ceph osd pool set volumes pg_autoscale_mode on The target_size_ratio is like the percentages customers input into pgcalc. Putting it directly into THT and skipping pgcalc is a better UX. The 16 in the first command was recommended as a hard coded value above while the name and target_size_ratio would be variables. The reason for 16 was based on Sage's blog [1] where he set 1, but 16 is the usual default and should work with replicated pools while not triggering the overdose protection. [1] https://ceph.io/rados/new-in-nautilus-pg-merging-and-autotuning/