[ovs-dev] [PATCH] [RFC] ovn: add start of ovn-nbctl

[Russell Bryant]

On 03/16/2015 05:24 PM, Gurucharan Shetty wrote:
> On Mon, Mar 16, 2015 at 2:21 PM, Ben Pfaff <blp at nicira.com> wrote:
>> On Mon, Mar 16, 2015 at 04:34:15PM -0400, Russell Bryant wrote:
>>> On 03/16/2015 12:08 PM, Russell Bryant wrote:
>>>> +Commands:\n\
>>>> +  add-lswitch               Create a logical switch\n\
>>>> +  list-lswitch              List configured logical switches\n\
>>>
>>> I ran into odd thing with "add-lswitch".  There isn't a friendly way to
>>> refer to logical switches you've created.  Perhaps it would make sense
>>> to just add a "name" to the schema for a logical switch.  Thoughts?
>>
>> I think that's reasonable.  We designed the schema with use by
>> programs in mind, not use by humans, and programs are happy enough
>> with UUIDs and so on, so we left out names.  But it doesn't cost much
>> to add a "name" column, so I guess we might as well do it.
> 
> If names are added, is there a requirement that the name be unique? If
> that is the case, would it not be hard for Neutron to maintain unique
> names across tenants?

I wouldn't expect Neutron to use the field.  My thinking for adding it
was just for convenience when doing dev and test from the command line.

Neutron networks map to an OVN logical switch [1].  Neutron uses a UUID
to uniquely identify a network, and I was proposing we stick that in as
an external:id.  I was also proposing sticking neutron's network name in
as an external:id, but that could be moved to the new "name" field.
Again though, the name is just for convenience and shouldn't need to be
unique.

On the command line (ovn-nbctl) side, I suppose it could accept either a
name or UUID and fail if a given name isn't unique.  That's typically
how OpenStack utilities behave.

[1]
http://git.openstack.org/cgit/stackforge/networking-ovn/tree/doc/source/data_model.rst

-- 
Russell Bryant