Ask Your Question

thin provisioned root disks

asked 2014-05-07 14:26:18 -0500

radclif gravatar image

updated 2014-05-08 04:17:54 -0500

Hi In our current setup we provision all vm instance disks thin. we do not use cinder or swift, and all disk images they are located under /var/lib/nova/instnaces

whilst this is good as it saves space I am now starting to worry about the possiblity of over commiting disk space on compute nodes.

How is this handled by Openstack does it keep track of these space claimed by each disk or the space actually used?

Also will openstack stop you from creating new instances if you exceed the theoretical limit (i.e there may be free disk space on the compute node but openstack will not create an instance because if all vm disks where to be filled up there would not be enough space)?

edit retag flag offensive close merge delete

1 answer

Sort by ยป oldest newest most voted

answered 2014-05-07 18:46:35 -0500

mpetason gravatar image

Look into the option:

disk_allocation_ratio=1.0 (FloatOpt) virtual disk to physical disk allocation ratio

Then you can look at nova scheduler to see how it calculates where to put new instances.

Basically you would want to configure the disk allocation ration to be 1:1 if you are worried about hitting the limit, or even leave some space in reserve. You could also achieve this with:

reserved_host_disk_mb =

The disk uses/available gets reported back to the scheduler from nova-compute running on the compute nodes. It updates what resources it still has available so that the scheduler can figure out where to put the next instance that gets launched. Depending on your deployment the calculation could be different, if you customize then you can add other attributes to look for when launching instances.

edit flag offensive delete link more


Short answer, nova-compute is always updating about available resources. It reports back CPU/RAM/Disk Space. If you look at the code that is pulling this data you can see how it's calculated/polled. It should report back with "disk used:" as you are expanding the disk size of your instances.

mpetason gravatar imagempetason ( 2014-05-07 18:57:34 -0500 )edit

I understand that an allocation ratio of 1:1 would mean more or less that no over provisioning would take place. however does a ratio of 1.0 mean unlimited disk allocation. I noticed as well that vcpu allocation defaults to 16:0 could you clear up how this ratio is worked out? or link an explination

radclif gravatar imageradclif ( 2014-05-08 04:23:17 -0500 )edit

You may have already seen this:

Little bit older but has some useful information:

mpetason gravatar imagempetason ( 2014-05-08 11:05:29 -0500 )edit

Get to know Ask OpenStack

Resources for moderators

Question Tools

1 follower


Asked: 2014-05-07 14:26:18 -0500

Seen: 765 times

Last updated: May 08 '14