[ovs-dev] [PATCH 0/4] Restore ovn-controller binding functionality.

Ryan Moats rmoats at us.ibm.com
Fri Jun 24 21:26:24 UTC 2016


Ben Pfaff <blp at ovn.org> wrote on 06/24/2016 04:16:14 PM:

> From: Ben Pfaff <blp at ovn.org>
> To: Ryan Moats/Omaha/IBM at IBMUS
> Cc: dev at openvswitch.org, Russell Bryant <russell at ovn.org>
> Date: 06/24/2016 04:16 PM
> Subject: Re: [PATCH 0/4] Restore ovn-controller binding functionality.
>
> On Fri, Jun 24, 2016 at 03:07:26PM -0500, Ryan Moats wrote:
> > Ben Pfaff <blp at ovn.org> wrote on 06/24/2016 02:59:30 PM:
> >
> > > From: Ben Pfaff <blp at ovn.org>
> > > To: Russell Bryant <russell at ovn.org>
> > > Cc: dev at openvswitch.org, Ryan Moats/Omaha/IBM at IBMUS
> > > Date: 06/24/2016 03:00 PM
> > > Subject: Re: [PATCH 0/4] Restore ovn-controller binding
functionality.
> > >
> > > On Fri, Jun 24, 2016 at 03:52:07PM -0400, Russell Bryant wrote:
> > > > "Convert binding_run to incremental processing" caused a regression
> > > > in port binding handling.  This was seen in OpenStack CI.  The last
> > > > commit here adds a test case that would have caught the issue in
OVN.
> > > >
> > > > There are 3 reverts as two later patches conflicted with reverting
> > > > patch 3/4.  This is just to show what has to be rolled back to get
> > > > to where the added test in patch 4/4 will pass.
> > > >
> > > > Ryan says he has a possible fix for the issue, so we should pursue
> > > > that before merging the reverts.
> > >
> > > This is interesting.  I expected some minor fallout, but I did not
> > > expect it to break severely.  Incremental processing is hard.
> > >
> > > Ryan, will your patch be ready soon?  At a glance, I do not see it
yet.
> > >
> >
> > It is hard, but I have a patch that looks hopeful - I've tested it
> > by hand, I'm running it through existing UT first and then I'll add
> > Russell's test and run it through UT again and then post it. I'd say
> > ETA of 30 minutes or less...
>
> I applied it before I saw these messages.
>
> If it doesn't do the job, and there's no obvious further fix, it's just
> one more revert, no big deal.
>

No worries - it does pass the new test case, which is good and we are
running
upstream rechecks to see if it does the trick.

BTW, I acked the new test case, with a proviso, so getting that in would be
nice, in case this doesn't fix things and we need to revert everything out.

RYan



More information about the dev mailing list