[openstack-dev] [nova] [os-vif] [neutron] Race in setting up linux bridge

Kevin Benton kevin at benton.pub
Mon Aug 29 12:29:30 UTC 2016


Sort of. The neutron agent code doesn't use os-vif because the os-vif devs
indicated that neutron's vif plugging code wasn't a use case they cared
about [1].

So if we did generalize os-vif to work with the neutron agents then it
would be two calling the same locked code. But at this point it's just two
versions of similar logic trying to do the same thing.

1. https://review.openstack.org/#/c/284209/

On Aug 29, 2016 05:56, "Sean Dague" <sean at dague.net> wrote:

> On 08/26/2016 05:23 PM, Armando M. wrote:
> > Folks,
> >
> > Today I spotted [1]. It turns out Neutron and Nova might be racing
> > trying to set up the bridge to provide VM with connectivity/dhcp. In the
> > observed failure mode, os-vif fails in [2].
> >
> > I suppose we might need to protect the bridge creation and make it
> > handle the potential exception. We would need a similar fix for Neutron
> > in [3].
> >
> > That said, knowing there is a looming deadline [4], I'd invite folks to
> > keep an eye on the bug.
> >
> > Many thanks,
> > Armando
> >
> > [1] https://bugs.launchpad.net/neutron/+bug/1617447
> > [2] https://github.com/openstack/os-vif/blob/master/vif_plug_
> linux_bridge/linux_net.py#L125
> > [3] http://git.openstack.org/cgit/openstack/neutron/tree/
> neutron/agent/linux/bridge_lib.py#n58
> > [4] http://lists.openstack.org/pipermail/openstack-dev/2016-
> August/102339.html
>
> Is this another issue where 2 processes are calling "locked" code, but
> from different projects, so the locks have no impact? We just went
> through this with os-brick and had to mitigate with a retry block.
>
> We probably need to make some strong rules about what locking can look
> like in these common libraries, because they have a lot of ported code
> that doesn't really work when communicating between 2 services.
>
>         -Sean
>
> --
> Sean Dague
> http://dague.net
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20160829/a9f67c4f/attachment.html>


More information about the OpenStack-dev mailing list