[c-nsp] IPv6 Migration with ISIS (was Route Reflector Design)

[Kristian Larsson]

On Fri, Jul 04, 2008 at 10:25:56AM -0400, Vinny Abello wrote:
> > -----Original Message-----
> > From: cisco-nsp-bounces at puck.nether.net [mailto:cisco-nsp-
> > bounces at puck.nether.net] On Behalf Of Mikael Abrahamsson
> > Sent: Friday, July 04, 2008 1:42 AM
> > To: cisco-nsp at puck.nether.net
> > Subject: Re: [c-nsp] IPv6 Migration with ISIS (was Route Reflector
> > Design)
> >
> > On Thu, 3 Jul 2008, Vinny Abello wrote:
> >
> > > While on this topic, if anyone has figured out a non-disruptive
> > strategy
> > > to deploying IPv6 in a core with a mix of Cisco and Foundry routers
> > > running ISIS, any pointers would be appreciated. Foundry currently
> >
> > We had multitopology problems between platforms/vendors as well, we
> > ended
> > up "solving" the issue by using OSPFv3 as IPv6 IGP (and ISIS for
> > IPv4/VPNv4), this gave us a completely different control plane for IPv6
> > and pretty much guaranteed to be non-intrusive to devices not running
> > IPv6
> > or needing the information.
> >
> > Multitopology ISIS is a great idea and I would really like to run it,
> > but
> > it just didn't work with our mix of platforms and vendors.
> 
> Thanks Mikael. I hadn't considered running OSPFv3 for IPv6. I'll have to see if that is a viable possibility in our network. Did you run into any challenges in doing this such as administrative distances of the routing protocols and things defaulting to using IPv6 instead of IPv4 or other unexpected results? In theory if you're only doing the IPv6 address family, I wouldn't expect any problems, but firsthand experience is always better than theory. :) By the way, what other vendor's or vendors' equipment were you working with besides Cisco where you had the same ISIS multi-topology challenges?

Apologies for a tad late answer, don't read my
nanog box to often... 

I was the one implementing this, so thought I'd
give you a few answers. We run RedBack SmartEdge
boxes with a variety of software, some supposedly 
supporting IPv6 others do not at all.
What happened with these boxes was that they threw
away ISIS LSPs which contained one or more v6 TLVs
resulting in that any IPv4 information in that LSP
was also thrown away. From what I've heard this
was the correct behaviour according to some early
ISIS standard, though I can not find any
mentioning of it in the current standards.

OSPFv3 is working very well. We are using
basically the same metric system as for ISIS IPv4
which makes administration quite easy. I say
basically for we have a few places where we use an
ISIS metric of > 65k and as OSPFv3 only support
16-bit link metrics while ISIS supports 24-bit
this becomes a slight annoyance.

"things defaulting to using ipv6 instead of
ipv4".. this sound more of a host-side symptom and
not one dependant upon your choice of IGP.

I expect to migrate our network to MT ISIS after
resolving all our issues with RedBack. The higher
administrative distance of ISIS allows us to
enable ISIS MT all over our network and run both
OSPFv3 and ISIS MT for IPv6 with no impact. We
would have plenty of time to compare the ISIS
database with the OSPFv3 one to make sure
everything looks good and then we can simply shut
down OSPFv3 and let ISIS take over.

If you want some more in-depth answers we can
prolly take it off-list.

Kind regards,
   Kristian.

-- 
Kristian Larsson                                        KLL-RIPE
Network Engineer / Internet Core        Tele2 / SWIPnet [AS1257]
+46 704 910401			              kll at spritelink.net