Open Stack

Hi All,

I am not sure how quantum is responsible for figuring out the
compute-host(as per your discussion).

All I knew is that its nova-scheduler whose job is to figure out which
compute node should be used for bringing up the instances. Correct me if I
am wrong.

@Piotr: Each nova-compute node keeps on updating nova-scheduler about the
amount of RAM + vCPUs it has. Basic nova-scheduler checks which host has
the maximum free RAM at that moment and launches the instance on that host.
In your case, all your instances would be coming on single compute node
since it would be having the max RAM even after the launch of 'n'
instances. The moment its free RAM becomes less than that of other
compute-node, next instance will be launched on the other node.

To know more about it, you can have a look at nova-scheduler and its
various configurations.
http://docs.openstack.org/trunk/openstack-compute/admin/content/ch_scheduling.html

-Regards
Rahul Sharma


On Fri, Sep 20, 2013 at 4:33 PM, Piotr Kopec <pkopec17 at gmail.com> wrote:

> Thanks for that Mahardhika,
>
> I believe you but I still don't see where this options is specified. Could
> you tell me where can I find it? (I'm working on Grizzly now)
> And another question: How can I configure quantum (because (correct me if
> i'm wrong) it is responsible for choosing compute node for instances) to
> distribute instances among compute nodes?
>
> Any links with some explenation about that issue will be appreciated.
> Unfortunately doc.openstack.org doesn't explain a lot about that.
>
> Regards
> Piotr
>
>
> 2013/9/20 Mahardhika Gilang <mahardika.gilang at andalabs.com>
>
>>  Tryin to answer here,
>> Instance will build on compute node that have nova-compute and kvm on it,
>> and determining which is compute with quantum configuration in
>> sql_connection section.
>>
>>
>> On 9/20/2013 4:41 PM, Piotr Kopec wrote:
>>
>>  Hello folks,
>>
>> I'm wondering how those compute nodes are working in practice. End user
>> will use Dashboard for building instances.
>> So I have deployed nova as follows:
>>
>> [root at ctlnode136 ~]# nova-manage service list
>> Binary           Host                                 Zone             Status     State Updated_At
>> nova-cert        ctlnode136                           internal         enabled    :-)   2013-09-20 09:34:04
>> nova-scheduler   ctlnode136                           internal         enabled    :-)   2013-09-20 09:34:04
>> nova-consoleauth ctlnode136                           internal         enabled    :-)   2013-09-20 09:34:04
>> nova-conductor   ctlnode136                           internal         enabled    :-)   2013-09-20 09:33:58
>> nova-compute     network                              nova             enabled    :-)   2013-09-20 09:34:01
>> nova-console     ctlnode136                           internal         enabled    :-)   2013-09-20 09:34:02
>> nova-compute     compute                              nova             enabled    :-)   2013-09-20 09:34:01
>> nova-console     compute                              internal         enabled    XXX   2013-09-20 09:32:20
>>
>> And each time when I‘m creating instance from Dashboard it’s admin panel
>> shows that those instances live on “compute” host. Not even single
>> instance, created from Dashboard, live on (in my case) “network” host.
>>
>> Is there any way to distribute equally among compute nodes instances
>> created from Dashboard?
>>
>> Thanks for replies.
>>
>> Piotr.
>>
>>
>> _______________________________________________
>> 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
>>
>>
>>
>> --
>> Regards,
>> Mahardhika Gilang
>>
>> *PT. Andalabs Technology *
>> Gedung Gravira
>> Jl. Cideng Barat no. 54
>> Jakarta Pusat 10150
>>
>> HP : 0852 139 55861
>> Email : mahardika.gilang at andalabs.com
>>
>
>
> _______________________________________________
> 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/20130920/5e1a396f/attachment.html>