Open Stack

With 4 physical cpu cores, you'll have 8 with HT so you likely don't need
anything managing overcommitting. But you also have memory contention in
your design. OpenStack facilitates resource limits to prevent the
contention from happening but does not get into managing vm priorities. The
assumption I presume is that high priority instances will be placed on high
priority hosts where overcommit ratios are safe for the VM's hosted on that
machine.

I know guys don't like to hear this when things are running but you need to
reassess what you're doing and rethink the 'why' factor to understand if
that level of manual management is really necessary. Remember this is a
cloud but your setup is basically complicated pet management. Add another
compute host and put your low priority instances on it rather than trying
to manage two SLA's yourself. Let OpenStack do it for you. ; )

//adam

All the best.

//adam

On Jan 12, 2017 9:17 AM, "Kostyantyn Volenbovskyi" <volenbovsky at yandex.ru>
wrote:

Hi,

I have provided the answer to that on https://ask.openstack.org/en/
question/101374 where you asked this question as well.
That elaborates slightly on 'There are facilities to allow one VM or
another to have CPU priority’ which James mentioned below.
For RAM such things as Memory Tuning https://libvirt.org/formatdomain.html#
elementsMemoryTuning is not supported
as well as memory ballooning (which might be part of answer I think) are
not supported in Nova.
And well, ‘don’t overcommit RAM’ is often a good principle to stick to.


BR,
Konstantin
> On Jan 12, 2017, at 1:33 PM, Ivan Derbenev <ivan.derbenev at tech-corps.com>
wrote:
>
> What are the facilities for cpu? That's the initial question - how can I
do it if this is possible
>
>
> Best Regards
> Tech-corps IT Engineer
> Ivan Derbenev
> Phone: +79633431774
>
> -----Original Message-----
> From: James Downs [mailto:egon at egon.cc]
> Sent: Thursday, January 12, 2017 12:56 AM
> To: Ivan Derbenev <ivan.derbenev at tech-corps.com>
> Cc: openstack at lists.openstack.org
> Subject: Re: [Openstack] Openstack+KVM+overcommit, VM priority
>
> On Wed, Jan 11, 2017 at 09:34:32PM +0000, Ivan Derbenev wrote:
>
>> if both vms start using all 64gb memory, both of them start using swap
>
> Don't overcommit RAM.
>
>> So, the question is - is it possible to prioritize 1st vm above 2nd? so
the second one will fail before the 1st, to leave maximum possible
perfomance to the most importan one?
>
> Do you mean CPU prioritization? There are facilities to allow one VM or
another to have CPU priority, but what, if a high priority VM wants RAM,
you want to OOM the other? That doesn't exist, AFAIK.
>
> Cheers,
> -j
>
> _______________________________________________
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/
openstack
> Post to     : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/
openstack


_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to     : openstack at lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20170112/03fe4981/attachment.html>