yifan writes:
> James Carlson wrote:
> There's another issue. The VRRP RFC states that the backup routers must 
> be configured to send the same Router Advertisement options as the 
> address owner. I think it makes no sense for VRRP to find some way to 
> sync the RA option setup between master and backups. Thus it should be 
> the router administrator who guarantees that the master and backups have 
> same setup on RA options. And ndpd.conf surely should be used for that.

Yes; that's what I think should be done.

> >For what it's worth, the VRRP I-D that you're reading
> >(vrrp-unified-spec) is weird in that it goes into extreme detail on
> >IPv6 issues (section 8.2.2), but completely ignores the identical IPv4
> >issues.
> >
> 
> Isn't section 8.1.2 for corresponding IPv4 issues?

It completely glosses over the issue.  It doesn't address IPv4 RDISC
at all (which is a subset of IPv6's RA mechanism), nor does it address
the more general issue of router-to-host communication or why some
things are "in" the list and some things are "out."

The actual issue here is that in order to be a competent router, you
need to continue advertising the information that the hosts on your
network need.  If that means advertising RIP, then that has to be
done.  If it means sending RDISC messages, then that's included.

What you must not do (for routing purposes) is impersonate the address
owner.  So, for example, responding to SNMP messages sent to the
virtual IP address in any way would be an error.  It'd be an error
because the non-owner doesn't have the same context, and the client
sending those requests would end up with a confused mess if a
fail-over happened in the middle of a MIB walk.

I think section 8.1 is essentially defective.

> >I think it's necessary to take it with a grain of salt.
> >
> >(I'll see if I can't get clarification from the author.)

There seems to be a big discussion about Accept_Mode on the
[EMAIL PROTECTED] mailing list.  If you're not monitoring that list now,
you probably should be.

> >That's one possibility.  Another might be to put the existing
> >IFF_NOLOCAL flag to some good use or invent a new interface flag.
> >
> 
> To make it work, in addition to using an interface flag to block all 
> traffic on link layer for backup state, there must be another filter on 
> IP layer to do the "forward only" job for master state.

Yes.

-- 
James Carlson, Solaris Networking              <[EMAIL PROTECTED]>
Sun Microsystems / 35 Network Drive        71.232W   Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757   42.496N   Fax +1 781 442 1677
_______________________________________________
networking-discuss mailing list
[email protected]

Reply via email to