Open Stack

Hi all,

I have found the problem of starting a nova-network on a controller node with single network node.

nova network-list
+--------------------------------------+----------+------------------+
| ID                                   | Label    | Cidr             |
+--------------------------------------+----------+------------------+
| f9eadb0b-1c0f-4c48-acdf-aa20987d22d1 | demo-net | 192.168.165.0/24 |

However, the "network" object has a "host" associated with it, which can be seen by directly quering MySQL table.  The network object above has the host associated with one the compute node.  If I start nova-network on any compute node, it would work properly, but if I start nova-network on controller node, the network initialization code would be skipped unless I reset the host to be the controller node.

I don't know if it is a feature or bug of the nova-network.  I hope that this information would be helpful for someone playing with nova-network.

Thank you.

Zhi-Wei Lu
IET-CR-Network Operations Center
University of California, Davis
(530) 752-0155

From: Zhi-Wei Lu [mailto:zwlu at ucdavis.edu]
Sent: Thursday, June 05, 2014 11:49 AM
To: openstack at lists.openstack.org
Subject: [Openstack] Question about debugging nova-network

Hi all,

I have posed a question a few weeks ago about debugging the nova-network issue, that I can not run the nova-network on the "controller" node (single flatdhcpmanger), I was able to run that properly on the compute nodes, however.  During the process of debugging, I understand the process of network much better, iptables, ebtable, dhcp, meta-data services all could cause problem.

I tracked it down in the /usr/lib/python2.6/site-packages/nova/network/l3.py

fixed_range = kwargs.get('fixed_range', False)
        networks = kwargs.get('networks', None)
        if not fixed_range and networks is not None:
            LOG.debug("I am here")
            for network in networks:
                LOG.debug("But not here")
                self.initialize_network(network['cidr'])

Therefore, nova-network skipped bridge configuration and dnsmasq, thus it is quite useless even though the main nova-network is running!  The "networks" must be different on the "controller node" and on "compute nodes".

How do I print the variables fixed_range and network using the openstack logging facilities?  I would like to get to the bottom of this issue.

Thank you very much for your help!

Zhi-Wei Lu
IET-CR-Network Operations Center
University of California, Davis
(530) 752-0155

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140606/a464e4a7/attachment.html>