[Openstack-operators] nova-conductor scale out

Kris G. Lindgren klindgren at godaddy.com
Tue Mar 15 15:42:28 UTC 2016


Yes.  Nova-conductor is rpc based so you can add as many servers as you need, and they will process messages from the conductor queue on rabbit, without any problems.  I would suggest also moving rabbitmq off on its own server as well.  As rabbitmq also chews up a significant amount of CPU as well.

___________________________________________________________________
Kris Lindgren
Senior Linux Systems Engineer
GoDaddy

From: Gustavo Randich <gustavo.randich at gmail.com<mailto:gustavo.randich at gmail.com>>
Date: Tuesday, March 15, 2016 at 9:38 AM
To: "Kris G. Lindgren" <klindgren at godaddy.com<mailto:klindgren at godaddy.com>>
Cc: David Medberry <openstack at medberry.net<mailto:openstack at medberry.net>>, "openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>" <openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>>
Subject: Re: [Openstack-operators] nova-conductor scale out

PD: 32 cores


On Tue, Mar 15, 2016 at 12:37 PM, Gustavo Randich <gustavo.randich at gmail.com<mailto:gustavo.randich at gmail.com>> wrote:
We are melting right now (rpc timeouts, rabbitmq connection timeouts, high load on controller, etc.): we are running 375 compute nodes, and only one controller (on vmware) on which we run rabbitmq + nova-conductor with 28 workers

So I can seamlessly add more controller nodes with more nova-conductor workers?


On Tue, Mar 15, 2016 at 11:59 AM, Kris G. Lindgren <klindgren at godaddy.com<mailto:klindgren at godaddy.com>> wrote:
We run cells, but when we reached about 250 hv in a cell we needed to add another cell api (went from 2 to 3) to help with the cpu load caused by novaconductor.  Nova-conductor was/is constantly crushing the cpu on those servers.

___________________________________________________________________
Kris Lindgren
Senior Linux Systems Engineer
GoDaddy

From: David Medberry <openstack at medberry.net<mailto:openstack at medberry.net>>
Date: Tuesday, March 15, 2016 at 8:54 AM
To: Gustavo Randich <gustavo.randich at gmail.com<mailto:gustavo.randich at gmail.com>>
Cc: "openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>" <openstack-operators at lists.openstack.org<mailto:openstack-operators at lists.openstack.org>>
Subject: Re: [Openstack-operators] nova-conductor scale out

How many compute nodes do you have (that is triggering your controller node limitations)?

We run nova-conductor on multiple control nodes. Each control node runs "N" conductors where N is basically the HyperThreaded CPU count.

On Tue, Mar 15, 2016 at 8:44 AM, Gustavo Randich <gustavo.randich at gmail.com<mailto:gustavo.randich at gmail.com>> wrote:
Hi,

Simple question: can I deploy nova-conductor across several servers? (Icehouse)

Because we are reaching a limit in our controller node....


_______________________________________________
OpenStack-operators mailing list
OpenStack-operators at lists.openstack.org<mailto:OpenStack-operators at lists.openstack.org>
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-operators




-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-operators/attachments/20160315/b7670efb/attachment.html>


More information about the OpenStack-operators mailing list