[ovs-dev] OVN OpenStack integration status

Russell Bryant rbryant at redhat.com
Wed May 13 18:17:57 UTC 2015


On 05/13/2015 01:28 PM, Gurucharan Shetty wrote:
> On Wed, May 13, 2015 at 10:19 AM, Russell Bryant <rbryant at redhat.com> wrote:
>> On 05/13/2015 11:57 AM, Justin Pettit wrote:
>>>
>>>> On May 13, 2015, at 6:48 AM, Russell Bryant <rbryant at redhat.com> wrote:
>>>>
>>>> There's some documentation for testing it with DevStack here:
>>>>
>>>> http://networking-ovn.readthedocs.org/en/latest/testing.html
>>>>
>>>> I also have a bunch of patches up for review to make multi-node devstack
>>>> work with OVN.  It's currently blocked by an OVN bug, though.  I'm sure
>>>> we'll have it sorted out this week, though.  That page will include the
>>>> multi-node test instructions once it's all merged.
>>>
>>> Hi, Russell.  Have you sent out anything about this OVN bug?  Ben and
>>> I have been traveling the last couple of weeks, so I just want to
>>> make sure that we didn't miss something.  Let me know if there's
>>> anything we can do.
>>
>> Not yet.  I had been working on my multi-node setup and just this week
>> got far enough where I think it should be working.  I'm having trouble
>> talking between hypervisors, though.  At the end of the day yesterday,
>> Gurucharan Shetty said that he now has the same problem with the latest
>> code in the ovn branch.  I was hoping to better understand what was
>> going on today so that I could either come up with a patch or a more
>> intelligent bug report.  :-)
>>
> 
> The problem at the looks of it is that when you have more than one VM
> in each hypervisor, the broadcast flow generated in table17 by
> ovn-controller misses broadcasting to one VM (if you just have one VM
> each, you can ping). The pipeline table in OVN_SB looks to have the
> correct entry (unless I am misreading it), but it does not get
> converted into openflow flows by ovn-controller properly.

Thanks for clarifying.  I was actually getting confused just now since
it seemed to be working with just 1 VM on each hypervisor while trying
to replicate it.  :-)

-- 
Russell Bryant



More information about the dev mailing list