Re: NetworkManager with multiple RA (radvd failover)

[Pavel Simerda <psimerda redhat com>]

----- Original Message -----
> From: "Anders Blomdell" <anders blomdell control lth se>
> To: "Pavel Simerda" <psimerda redhat com>, "Dan Williams" <dcbw redhat com>
> Cc: networkmanager-list gnome org
> Sent: Thursday, January 2, 2014 6:38:48 PM
> Subject: Re: NetworkManager with multiple RA (radvd failover)
>
> On 2013-12-18 08:48, Anders Blomdell wrote:
> > On 2013-12-18 00:00, Pavel Simerda wrote:
> > > ----- Original Message -----
> > > > From: "Dan Williams" <dcbw redhat com>
> > > > To: "Anders Blomdell" <anders blomdell control lth se>
> > > > Cc: networkmanager-list gnome org
> > > > Sent: Tuesday, December 17, 2013 11:58:34 PM
> > > > Subject: Re: NetworkManager with multiple RA (radvd failover)
> > > >
> > > > On Tue, 2013-12-17 at 13:26 +0100, Anders Blomdell wrote:
> > > > > Hi,
> > > > >
> > > > > I'm trying to implement RA failover at my site, but NetworkManager grabs
> > > > > the
> > > > > first RA it sees to set a (static) default route with metric 1, and then
> > > > > it
> > > > > sits there even when the lifetime of this route has expired (and no more
> > > > > RA's
> > > > > from that host). Is this a bug or some feature tham I'm overlooking.
> > > >
> > > > Which version of NetworkManager do you have?  I  know there were some
> > > > lifetime issues with previous versions of NM, but it would certainly be
> > > > a bug if NetworkManager continued to leave a default route up when the
> > > > router's lifetime had expired.
> > >
> > > The metric one suggest it's a pre-libndp version. At that time the route
> > > management was rather difficult.
> > NetworkManager-0.9.8.8-2.fc19.x86_64
> >
> > Time to switch to fc20 now, will be back when this is done.
> Ok, when running NetworkManager-0.9.9.0-22.git20131003.fc20.x86_64 default
> route changes
> when the lifetime of a (dead) route has expired. Unfortunately that doesn't
> help much
> since old routes to specific hosts seems to be stuck in the kernel (at least
> when using ping6),
> like this:
>
>   2a00:1450:4005:808::1011 via fe80::64:1ff:fe00:2 dev eth0  proto static
>   metric 1
>   2a00:1450:4005:808::1013 via fe80::13:10ff:fe07:1158 dev eth0  proto static
>   metric 1
>   2a00:1450:4005:808::1014 via fe80::13:12ff:fe10:1311 dev eth0  proto static
>   metric 1
>   default via fe80::13:10ff:fe07:1158 dev eth0  proto static  metric 1024
>
>
> Where 2a00:1450:400f:800::XXXX are different incarnations of www.google.com.
>
> Could keepalived (VRRP) be a solution? What is everyone else using?

I don't think we need to make things complicated. Metric one routes are typically added by older versions of 
NetworkManager. Also, kernel routes are usually marked so that you can distinguish them. I would be happy if 
you could start a bugzilla ticket (bugzilla.gnome.org) to track this issue so that we can maintain some 
resources (like log output, etc). Please also describe the proces from a clean boot status.

Thanks a lot!

Pavel


> /Anders
>
> --
> Anders Blomdell                  Email: anders blomdell control lth se
> Department of Automatic Control
> Lund University                  Phone:    +46 46 222 4625
> P.O. Box 118                     Fax:      +46 46 138118
> SE-221 00 Lund, Sweden