I'd be happy with the out-of-tree option, especially if that gives folks such as myself the opportunity to breathe some life back into it and at the very least validate compatibility and have better visibility of the driver's viability in releases post-Zed. -- -Nick On 16 Jan 2025 at 22:03:48, Brian Haley <haleyb.dev@gmail.com> wrote:
Yes, one of the issues is eventlet - we are in the process of migrating all of neutron away from it, and Linuxbridge was one of the open issues. Not doing that will allow us to work on all the other parts.
Could Linuxbridge be moved to an out-of-tree driver like networking-ovn was? I would think that is possible, yes. There is a process for creating such projects, just don't have a link.
-Brian
On 1/16/25 4:47 PM, Dmitriy Rabotyagov wrote:
From what I can recall, the biggest hassle at the moment with the
linuxbridge driver - is that it does rely on the eventlet library.
There's a proposed community goal to replace eventlet [1] , though the
replacement process is time consuming and not really straightforward. So
no active contributors to are ready to takeover this migration process.
While it could be not the only reason, it's one I heard as close to be
one of the most problematic ones during some discussion.
But I'd prefer giving the word to Neutron team here, as they obviously
have better overview here.
And yes, moving the plugin to a separate repo is always an option, and
it can be maintained by a different subset of people, while most of the
content being copied over from existing one.
[1]
https://governance.openstack.org/tc/goals/proposed/remove-eventlet.html
<https://governance.openstack.org/tc/goals/proposed/remove-eventlet.html>
On Thu, 16 Jan 2025, 22:26 Nick Jones, <nick@dischord.org
<mailto:nick@dischord.org>> wrote:
Thanks Mohammed. I'm super out of the loop on governance, structure
etc. so what would that look like in practice - would that give us a
plugin that could be referenced, and it'd be maintained with its own
set of tests etc., presumably copied over / out from the existing
Neutron codebase?
--
-Nick
On 16 Jan 2025 at 21:08:48, Mohammed Naser <mnaser@vexxhost.com
<mailto:mnaser@vexxhost.com>> wrote:
I think the best way to go about this would be if someone cares
enough to maintain it for them to create networking-linuxbridge in
the same way that networking-ovn existed before and they can
continue to maintain it there (possibly under the |x| namespace).
This will remove the burden of continuing to maintain it and those
who want to continue to run and maintain it can do it there.
------------------------------------------------------------------------
*From:* Nick Jones <nick@dischord.org <mailto:nick@dischord.org>>
*Sent:* January 16, 2025 4:04 PM
*To:* Brian Haley <haleyb.dev@gmail.com <mailto:haleyb.dev@gmail.com
*Cc:* openstack-discuss@lists.openstack.org
<mailto:openstack-discuss@lists.openstack.org>
<openstack-discuss@lists.openstack.org
<mailto:openstack-discuss@lists.openstack.org>>
*Subject:* Re: [neutron][upgrades] Linuxbridge driver removal and
migration
You don't often get email from nick@dischord.org
<mailto:nick@dischord.org>. Learn why this is important
Playing devil's advocate slightly, and as an operator who has had
one eye on migration (with no good answers really since it's quite
a jump - although I appreciate the efforts that have gone in to
showing that it can be done in principal), what would be entailed
to reinstate the LinuxBridge driver? It's been rock solid in my
case, and its simplicity is really appreciated.
--
-Nick
On 14 Jan 2025 at 19:40:26, Brian Haley <haleyb.dev@gmail.com
<mailto:haleyb.dev@gmail.com>> wrote:
Hi Stackers,
In 2022, during the Zed cycle, the Neutron team took a step at
marking
some items as experimental. One of those was the Linuxbridge
driver. The
reason was due to it being unmaintained for a number of cycles,
and no
one was able to take on its ownership. The implication of this
decision
was that it was consider deprecated, even if not given an official
warning message.
In doing this, all gate jobs were moved into the experimental queue,
feature-parity with other drivers was deemed unnecessary, and bug
fixing
was considered best-effort. Since then, both major distros have
stopped
supporting new deployments using Linuxbridge, and have been mainly
focused on OVN.
So what does this mean? We (the neutron cores) have come to the
conclusion that it is time to take the next step and remove the code
from the tree [1]. After 5 cycles of being considered experimental
without any investment it just makes sense. We plan on doing this
in the
current Epoxy cycle.
We realize that there are still deployments using Linuxbridge, and
hopefully they have been looking at migration strategies since we
originally marked the code experimental. This email is another
push to
the user community to start looking at this, especially if you are
planning any new deployments. There has been at least one migration
guide written at [2] (thanks Jim!), and at this point I would
encourage
those still using Linuxbridge to look at it and start asking any
questions they have so we have the chance to make it better and
can make
that process easier.
Thanks for reading,
-Brian, Neutron PTL