[Rdo-list] Trying out Neutron Quickstart running into issues with netns (l2 agent and dhcp agent)

[Perry Myers]

On 08/13/2013 08:10 PM, Perry Myers wrote:
> On 08/13/2013 08:09 PM, Perry Myers wrote:
>> On 08/04/2013 09:57 AM, Perry Myers wrote:
>>> Hi,
>>>
>>> I followed the instructions at:
>>> http://openstack.redhat.com/Neutron-Quickstart
>>> http://openstack.redhat.com/Running_an_instance_with_Neutron
>>
>> Ok, I wanted to close the loop on this thread with some things I found.
>>
>> It appears that the issues I was seeing had to do specifically with the
>> 114.openstack kernel in the RDO Grizzly repos.  This kernel had issues
>> with being able to change MAC addresses with an interface that is in the
>> up state.
>>
>> A new kernel was uploaded today 114.openstack.gre.2 and rkukura sent an
>> announcement out about it, as it enables gre tunnel support.
>>
>> A side effect of this new kernel is that it fixed the bugs with changing
>> MAC addresses on interfaces that are up.
>>
>> So, with this new kernel, I'm now able to successfully get RDO Grizzly
>> on RHEL 6.4 with the 114.openstack.gre.2 kernel running.  I can boot a
>> Cirros VM, get an ip address, get external connectivity from the VM and
>> also ssh/ping it via the floating IP.
>>
>> So... all in all, success!
>>
>> Now, the next step is doing a multi host setup.  This is where the new
>> support for gre tunnels will be handy, since moving from 1 to many nodes
>> requires either VLAN support or tunnels, and since I don't have an
>> intelligent switch, gre tunnels it is.
> 
> Oh, one thing to note... I did all of the above on hosts (VMs) where
> NetworkManager was removed.  I should try on hosts where NM is enabled
> to see if I still have success...

Ok, semi-encouraging results.  Here's what I did.

Test environment:
* RDO Grizzly
* 114.openstack.gre.2 kernel on RHEL 6.4.z
* single VM for an allinone install with eth0 as the NIC on the VM

Tests run were:

(a) booting cirros instance
(b) check for dhcp lease
(c) outbound connectivity from guest
(d) ping/ssh to FIP

* NM enabled and eth0 on the host with NM_CONTROLLED=yes
  a+ b+ c+ d-

* I reran the base test above to double check that I could repeat the
  failure
  a+ b+ c+ d+

So... inconsistent results.  I'd like to see what other folks find, and
if anyone else has issues with NM being enabled.  As of right now, I
don't have enough data to say that it is problematic or not.

Perry