[Openstack] Neutron GRE + Flat question

Randy amps at djlab.com
Tue Jun 10 19:30:26 UTC 2014


 

Thanks for the reply, 

So, it *should* be possible to pull in public network(s) over flat/vlan
to the VMs WITHOUT NAT, while still using GRE for private networking
(two NIC per VM... one public with no NAT, the other private via GRE)? 

I guess I'm trying to see if this is even possible - I think it should
be, but XenServer is making this more complicated and there are no
decent ML2-based XenServer examples out there. 

I'm still not sure how to expose a trunk port to a guest to use VLAN
networking on the network guest... 

~Randy 

On 06/10/2014 1:32 pm, Martinx - ジェームズ wrote: 

> Hello Randy, 
> 
> Just for the record, I'm using Flat Networks (very similar with VLAN Provider Networks), with a working Metadata connectivity... Instances get a route to the Metadata network through DHCP, in my case... 
> 
> Flat Networks, AFAIK, requires a "dedicated bridge map", using different physical ethernet from the one used by your GRE/VXLAN tunnels... 
> 
> Here is a working howto about Flat Networks with Metadata: https://gist.github.com/tmartinx/9177697 [4] - Easy to convert it to VLAN Provider Networks... 
> 
> Cheers! 
> 
> On 10 June 2014 09:17, Randy <amps at djlab.com> wrote:
> 
>> I've got the 3-node Icehouse/Juno setup using Neutron (ml2/GRE) networking on Xenserver, like this:
>> 
>> http://docs.openstack.org/trunk/install-guide/install/apt/content/basics-networking-neutron.html [1]
>> 
>> Everything works fine with the floating IPs and all...but now I want to introduce a flat network (with external router) that doesn't require NAT or floating IPs. Like this:
>> 
>> http://docs.openstack.org/havana/install-guide/install/apt/content/section_use-cases-mixed.html [2]
>> 
>> Using the flat network, DHCP works but there's no metadata or connectivity in or out of the VMs.
>> 
>> ACTUAL QUESTION: If we introduce the flat network to the network node on a dedicated interface, will it tunnel the flat network over GRE to each compute node? Or do we need to physically connect and bridge map each flat network on every compute node separately?
>> 
>> -- 
>> ~Randy
>> 
>> _______________________________________________
>> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack [3]
>> Post to : openstack at lists.openstack.org
>> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack [3]

-- 
~Randy
 

Links:
------
[1]
http://docs.openstack.org/trunk/install-guide/install/apt/content/basics-networking-neutron.html
[2]
http://docs.openstack.org/havana/install-guide/install/apt/content/section_use-cases-mixed.html
[3] http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
[4] https://gist.github.com/tmartinx/9177697
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack/attachments/20140610/f4f7ec67/attachment.html>


More information about the Openstack mailing list