Open Stack

Hi, I'm using cinder with lvm (mitaka release). Actually all works well,
but I noticed that when I create a new bootable volume from an image, the
storage backend has an intensive cpu/disk usage with a load around 12. I
think this happens because cinder use dd with oflag=direct during the copy.
I understand the reason of the oflag parameter, but is there some way to
solve the issue? The problem can also be related to the backend storage
however. I'm using lvm but the volume group is on a san and cinder-volume
uses iscsi for the access with a 10gb connection. The san is a drbd cluster
with tgtd. Please help me to analyze if I can fix something in cinder or I
should focus on the san.

Thank you
Marco
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20170124/f27cc178/attachment.html>