High Load on the Ceph Cluster Being Used for Cinder

Problem

You are observing increased IOPS to the Ceph clusters especially for the Cinder pool.

Environment

• Platform9 Managed OpenStack - All Versions

Cause

Cinder queries the Ceph cluster periodically to gather data regarding the Cinder backends, volumes and the snapshots present on the Cinder__pool.

Resolution

1. Add the rbd_exclusive_cinder_pool property to the cinder_override.conf file across all your Cinder hosts. This property needs to be added to the backend configuration stanza.

pf9@test ~ $ cat /opt/pf9/etc/pf9-cindervolume-base/conf.d/cinder_override.conf

[DEFAULT]

enabled_backends  =  ceph

glance_api_version  =  2

​

[ceph]

volume_backend_name  =  ceph1

volume_driver  =  cinder.volume.drivers.rbd.RBDDriver

rbd_exclusive_cinder_pool = true

2. Restart the pf9-cindervolume-base service for the changes to take effect

pf9@test ~ $ systemctl restart pf9-cindervolume-base

Additional Information

The value for the property is going to be set to True by default starting with the Wallaby release as per the OpenStack release notes.