Hi Mark,
As discussed on the v14 thread [0] I went ahead and applied this with
the change to sync NB -> SB uuids folded in [1].
On 8/13/25 7:20 PM, Mark Michelson via dev wrote:
> Prior to this commit, if you wanted to find the corresponding northbound
> UUID for a southbound Logical Datapath, you could find it in one of two
> places:
>
> For logical switches, it was in external-ids:logical-switch.
> For logical routers, it was in external-ids:logical-router.
>
> With this commit, we are separating the type and UUID into separate
> fields. This way, no matter the type of the datapath, you can find the
> UUID. And you can find the type of the datapath without having to
> potentially check multiple external-ids keys to do so.
>
> These fields are going to be used pretty heavily by northd in upcoming
> patches, so instead of making them external-ids, these are now
> full-fledged columns on the southbound Datapath_Binding. The UUID of the
> northbound logical datapath is in a column called "nb_uuid" and the type
> of the logical datapath is stored in an enumerated column called "type".
>
I rephrased a bit this part to match what we actually do
(sync NB -> SB UUIDs):
Datapath_Binding: Separate type column and sync NB.UUID to SB.
Prior to this commit, if you wanted to find the corresponding northbound
UUID for a southbound Logical Datapath, you could find it in one of two
places:
For logical switches, it was in external-ids:logical-switch.
For logical routers, it was in external-ids:logical-router.
With this commit, ovn-northd will be using the Northbound UUID of the
logical switch or router as UUID when creating the corresponding
Southbound Datapath_Binding. This way, no matter the type of the
datapath, you can find the Northbound record by UUID. And you can
find the type of the datapath without having to potentially check
multiple external-ids keys to do so.
The "type" field is going to be used pretty heavily by northd in upcoming
patches, so instead of making it an external-id, it is now a
full-fledged column on the southbound Datapath_Binding. The type
of the logical datapath is stored in a column called "type".
I hope that's fine with you.
> Note that there is a seemingly-unrelated change in the check packet
> length test in tests/ovn.at. This test started failing because its use
> of `grep` was picking up the new "nb_uuid" column accidentally. The
> change to the test uses `fetch_column` since it is more precise.
I reverted this part of the patch because it's not needed anymore if
we don't add a nb_uuid column.
>
> Signed-off-by: Mark Michelson <mmich...@redhat.com>
> Acked-by: Ales Musil <amu...@redhat.com>
> ---
As mentioned in [0], we still have time to polish this (if needed) before
the actual 25.09.0 release is cut.
Thanks again, everyone, for the hard work on getting this done!
Regards,
Dumitru
[0] https://mail.openvswitch.org/pipermail/ovs-dev/2025-August/425338.html
[1] https://github.com/dceara/ovn/commit/b4345db
_______________________________________________
dev mailing list
d...@openvswitch.org
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
