[Openstack] Unable to boot instances when multiple Neutron networks present

Brian Haley brian.haley at hp.com
Thu Sep 3 16:22:06 UTC 2015


On 09/02/2015 02:13 PM, Matt Dupre (projectcalico.org) wrote:
> Hi all,
>
> We've been seeing a problem where we can boot instances when there's one
> Neutron network configured, but not if a second is configured (even
> though the instance isn't using that network!).

Did you specify both networks at boot, or none?

> The symptom is an ERROR status in the dashboard, and a
> "MessagingTimeout: Timed out waiting for reply to message ID ..."
> traceback in the nova-api log file inside a "validate_networks" call.
> There are no other errors in any of the other log files, and no evidence
> of anything unusual at all in the Neutron log file, even down to DEBUG
> level.
>
> Deleting the second network clears the problem: instances boot
> successfully again.  I've included the nova-api.log traceback below -
> we're using OpenStack Juno.
>
> Has anybody seen this before, or does anyone have any suggestions for
> what a good next debugging step is?  We were thinking of writing some
> additional log statements into Nova and Neutron around the network
> creation and validation areas - does this sound sensible?

I can't comment on the exact error you're seeing, but I know the current 
behavior if you don't specify a network, but have more than one, is to error, 
but it should be something like, "Multiple possible networks found, use a 
Network ID to be more specific."  See 
https://bugs.launchpad.net/nova/+bug/1267723 for a little more info and a link 
to the patch that added the check for >1 network.

-Brian




More information about the Openstack mailing list