[PATCH] Preserve MED from eBGP peers

[Trisha Biswas]

Hi Ondrej,

Thanks for getting back. Yes, this would only affect routes that are
propagated to eBGP peers. So when a route is propagated from a true eBGP
peer to an internal eBGP peer, then the MED value would be preserved. This
is useful for networks where you have eBGP peers and would still like to
preserve the MED attribute beyond AS boundaries.

Although we are not aware of Cisco/Juniper having this option, this RFC on
MED considerations mentions such behavior:
https://tools.ietf.org/html/rfc4451#section-3.3. Thanks for the suggestion
on using export filters to set this attribute. For our use case, it feels
simpler to enable a knob for this, but we will test using export filters
for this.

Thanks,
Trisha
--

*Trisha Biswas* | Sr. Software Engineer, Network Systems
fastly.com | @fastly <https://twitter.com/fastly> | LinkedIn
<http://www.linkedin.com/company/fastly>


On Wed, Dec 9, 2020 at 8:03 PM Ondrej Zajicek <santiago at crfreenet.org>
wrote:

> On Tue, Dec 08, 2020 at 09:02:15AM -0800, Trisha Biswas wrote:
> > Hi all,
> >
> > The attached patch helps preserve MED values to/from eBGP peers. When
> > propagating a route from a true eBGP peer to an internal peer, this
> avoids
> > resetting the MED attribute to allow propagation of MED to/from external
> > peers. This is enabled via a configuration line: "preserve med", which
> goes
> > under BGP configs. Without the configuration line this would be a no-op.
>
> Hi
>
> I am bit hesitant with this patch. Seems to me that it is something that
> could be easily done with 'bgp_med = bgp_med;' export filter, but that
> may be tricky for users that are not thoroughly familiar with bgp_med
> behavior.
>
> Also, your description seems confusing. When a route is propagated from
> EBGP to an internal peer, then MED is preserved regardless of this.
> It is just propagation to EBGP peer when original MED is stripped.
>
> What is a use case for this option? Does other BGP implementations
> (Juniper, Cisco) have some similar option?
>
> --
> Elen sila lumenn' omentielvo
>
> Ondrej 'Santiago' Zajicek (email: santiago at crfreenet.org)
> OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net)
> "To err is human -- to blame it on a computer is even more so."
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://trubka.network.cz/pipermail/bird-users/attachments/20201211/3d08349e/attachment.htm>