Win2000 RRAS Dialin Client Routing Problem

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

The system is Windows 2000 Server, Service Pack 4, with most current updates applied

Among the systems functions is to provide dialin users access to the internet through a gateway at 192.168.0.1. This was working till about a week ago when the system was rebooted

The system has two NICs

Internet: xx.yy.zz.12, Mask 255.255.255.240 gateway xx.yy.zz.
Intranet: 192.168.0.6 Mask 255.255.255.0, gateway 192.168.0.

The dialin client systems are assigned addresses dynamically by RRAS in the range 192.168.253.3-192.168.253.255
Witn no routes statically defined in RRAS, attempting to PING/TRACERT systems on the Internet from the client systems produce timeouts. If a static route is defined in RRAS (e.g. Internal, 0.0.0.0, Mask 0.0.0.0, Gateway 192.168.0.1 or .6) we get the "192.168.253.2: Destination in unreachable" message

This configuration was working until the system was rebooted

What has happened? Ideas and thoughts appreciated

- Bob
 
Ok, I think that I can cancel this one, I puzzled it out

It is, IMHO, a bug in the way that the different interfaces deal with their properties during startup

It appears that if more than one LAN interface has a default route, the first one encountered during startup becomes the system wide default route. The others are ignored. However, there appears to be no way to deal with alternate adapters, nor for that matter is an error message placed in the event log about encountering a second default route

Needless to say, clearing the offending default fixed the problemn

- Bob
 
That is correct. It is a bit confusing, because it appears that you can
set a default gateway for each NIC in the TCP/IP settings. But a machine can
only have one default route. If you accidently set more than one default
route with the same priority, one of them will be chosen at random to act as
the default.

Bob Gezelter said:
Ok, I think that I can cancel this one, I puzzled it out.

It is, IMHO, a bug in the way that the different interfaces deal with
Click to expand...
their properties during startup.
It appears that if more than one LAN interface has a default route, the
Click to expand...
first one encountered during startup becomes the system wide default route.
The others are ignored. However, there appears to be no way to deal with
alternate adapters, nor for that matter is an error message placed in the
event log about encountering a second default route.
 
Not a bug. As below, there are cases where that configuration is what the
user wants.

There can only be one *active* default route in the PC's routing table.
Cisco routers (and others) will in certain cases load balance traffic
between two equivalent routes, I don't think we'd want that behaviour on
end-hosts.
Setting two default gateways isn't always a misconfiguration so an "error"
report would be wrong--though a "warning" might be useful. There are
instances where having two default gateways on two different links is
correct. I'd admit that most situations people are configuring today are
not in that situation though. I /believe/ Windows Server 2003 displays a
"warning" dialog when one enters a default gateway when an existing NIC
already has one, which is perhaps the best available solution.

I think the term "default gateway" is slightly unfortunate. In full it is
really, "address of a router that should be used to handle traffic to
destinations we don't have better routes for", or "router to use as a last
resort" or "router to use as a default when there is no better route", or
something of those forms. I don't know whether a better term might lead to
less confusion over multiple default gateways, or whether we all expect to
have to always fill in the default gateway field when we configure any
interface.

The case where more than one default gateways are configured mostly only
occurs when the two gateways are equivalent, e.g. a PC connected (for some
reason!) to two separate links on the same fully connected corporate
network. If one of the routers fail then the other default gateway will at
that time then be selected by Windows. Also as the two gateways are
equivalent it doesn't matter which is used, so the arbitrary selection of
which to use at start-up is fine. I have configured systems in this mode
myself.
That is correct. It is a bit confusing, because it appears that
you can set a default gateway for each NIC in the TCP/IP settings.
But a machine can only have one default route. If you accidently set
more than one default route with the same priority, one of them will
be chosen at random to act as the default.
Click to expand...
(comments as above).

Getting slightly off-topic...

Does anyone know of clear documentation on what happens when two equivalent
routes (e.g. default routes i.e. 0.0.0.0/0) have different priorities? I
haven't seen any myself and would welcome some. Does dead-gateway detection
operate? If so, is the higher-priority route always chosen at start-up (I
presume so).

Also is there clear documentation on whether dead gateway detection operates
on equivalent non default routes? tcpip2000.doc always talks in terms of
"the default gateway", but I've seen hints elsewhere that it apply to other
routes types too.
 
I am sure all this is documented somewhere, but I can't give you a
reference. All the references in MSDN library seem to refer to NT. There
were a few postings on this topic recently in the windows.server.networking
newsgroup, and Dusty Harper from Microsoft answered some of the issues you
raise. Have a browse through there.

From memory, I think dead gateway detection only operates if the default
routes are using the same priority. And it is only triggered by TCP traffic
(UDP is ignored). I also believe it only applies to default routes.
 
Bill Grant said:
I am sure all this is documented somewhere, but I can't give you a
reference. All the references in MSDN library seem to refer to NT.
There were a few postings on this topic recently in the
windows.server.networking newsgroup, and Dusty Harper from Microsoft
answered some of the issues you raise. Have a browse through there.

From memory, I think dead gateway detection only operates if the
default routes are using the same priority. And it is only triggered
by TCP traffic (UDP is ignored). I also believe it only applies to
default routes.
Click to expand...
Thanks Bill, I'll have a look there.
 
Back
Top