← Back to team overview

openstack team mailing list archive

Re: Multiple compute hosts with HA configuration - compute node doesn't bring ip from dnsmasq-dhcp

 

Thanks for the reply Dan!
Yes, I am running with the option --multi_host=T on both controller and compute nodes [I followed the same URL that you mentioned in your email]
Controller node is running dnsmasq-dhcp correctly (There are 2 instances of it). If the VM is launched on the controller, then, everything works fine. It gets  ip leased from dnsmasq-dhcp server. However, if the VM is launched on the compute node (that runs nova-network/nova-compute/nova-api), then, it is getting ip leased from the corporate dhcp server.
Thanks,
-Jay

From: Dan Wendlandt <dan@xxxxxxxxxx>
To: Jayashree(Jay) Beltur <jaybeltur@xxxxxxxxx> 
Cc: "openstack@xxxxxxxxxxxxxxxxxxx" <openstack@xxxxxxxxxxxxxxxxxxx> 
Sent: Tuesday, February 14, 2012 9:55 AM
Subject: Re: [Openstack] Multiple compute hosts with HA configuration - compute node doesn't bring ip from dnsmasq-dhcp

Hi Jay, 

If you are running nova-network in a mode that performs DHCP, you should be sure that the network Nova is connecting VMs to does not have your corporate DHCP server on it.  Having two conflicting DHCP servers on the same network will give unpredictable results.  What may be happening is a "race" between the two DHCP servers.  

The second question is whether dnsmasq is actually running.  You didn't say that you were using --multi_host=True, so by default the network may have been assigned to either of the two network-nodes you created.  This would mean that dnsmasq might be running on either of the two servers running nova-network (either the controller server or compute server).  Can you check if it is running on the other server?  If it is not running on either, we'd probably have to see nova-network logs.   For more info on the --multi_host flag, see: http://docs.openstack.org/trunk/openstack-compute/admin/content/existing-ha-networking-options.html

Dan

On Tue, Feb 14, 2012 at 9:31 AM, Jayashree(Jay) Beltur <jaybeltur@xxxxxxxxx> wrote: 
Hello,
>I have a two node cluster running on Ubuntu 11.04 (natty) with the Diablo release from ppa:openstack- release/ 2011.3
>[2011.3 (2011.3-nova-milestone-tarball:tarmac-20110922115702-k9nkvxqzhj130av2)]
>
>Controller node  runs nova-compute, nova-volume, nova-network, nova-api - 2 NICs
>Compute node runs nova-compute, nova-network and nova_api - 2 NICs
> 
>I am using flatDHCPManager.
>
>When I launch on instance on the compute node, the VM is getting ip from the company dhcp server. It is not getting IP from the dnsmasq-dhcp server.
>Also, I do not see dnsmasq-dhcp up on the compute node.
>
>In the console log, I see that 
> 
>Sending discover...Sending select for 16.89.26.95...Lease of 16.89.26.95 obtained, lease time 1296000starting DHCP forEthernet interface eth0  [  OK  ]cloud-setup: checking http://169.254.169.254/2009-04-04/meta-data/instance-idwget: can't connect to remote host (169.254.169.254): No route to hostcloud-setup: failed 1/30: up 2.04. request failedwget: can't connect to remote host (169.254.169.254): No route to hostcloud-setup: failed 2/30: up 6.05. request failed
>How to bring up dnsmasq-dhcp on the compute node?
>
>Any help is appreciated.
>
>Thanks,
>-Jay
> 
>_______________________________________________
>Mailing list: https://launchpad.net/~openstack
>Post to     : openstack@xxxxxxxxxxxxxxxxxxx
>Unsubscribe : https://launchpad.net/~openstack
>More help   : https://help.launchpad.net/ListHelp
>
>-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~Dan Wendlandt  
Nicira Networks: www.nicira.com 
twitter: danwendlandt~~~~~~~~~~~~~~~~~~~~~~~~~~~

References