[one-users] balancing VM placement

[Mark Devlin]

We're using the (mostly) default scheduler settings.  policy=1, MAX_HOST=1, etc.  My question came about because we had an empty host and instantiated a number of VMs, and they all landed on the empty host.  I have no doubt we have something funky in our setup.  I'll have to experiment with the scheduler more.

Thanks for the quick reply, Carlos.



On Aug 1, 2014, at 7:47 AM, Carlos Martín Sánchez <cmartin at opennebula.org<mailto:cmartin at opennebula.org>> wrote:

Hi,

On Thu, Jul 31, 2014 at 7:47 PM, Mark Devlin <mdevlin at intelius.com<mailto:mdevlin at intelius.com>> wrote:
Greetings,

Is there a way, using 4.4 or newer, for the scheduler to place VMs based on their names?

No, but you can set a specific SCHED_REQUIREMENTS and _RANK in the Template [1]

 Basically I want to spread a type of VM across as many hosts as possible, as long as the hosts have available resources for it.

For example, if I have four hosts that have VMs and one that's empty, I don't want the five (web1-web5) VMs I'm about about to instantiate to all end up on that empty host.  I want one on each host.  However, if I start three more (web6-web8), it should place based on available resources.

Hopefully that makes sense.  Thank you

-Mark

That's how the scheduler works by default [2]. The policy is Striping, and MAX_HOST is set to 1. The striping policy will try to send the VMs to the host with less running VMs, but because of max_host = 1, only one VM will be sent to the empty host in that scheduling cycle.

If you change the default policy, use SCHED_RANK = "- RUNNING_VMS" in your Template.

Regards


[1] http://docs.opennebula.org/4.6/user/references/template.html#placement-section
[2] http://docs.opennebula.org/4.6/administration/references/schg.html
--
Carlos Martín, MSc
Project Engineer
OpenNebula - Flexible Enterprise Cloud Made Simple
www.OpenNebula.org<http://www.opennebula.org/> | cmartin at opennebula.org<mailto:cmartin at opennebula.org> | @OpenNebula<http://twitter.com/opennebula><mailto:cmartin at opennebula.org>





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opennebula.org/pipermail/users-opennebula.org/attachments/20140801/c2bff312/attachment.htm>