Re: Problem with IP Masquerade + routed internal network (pretty newbie question)

From: David Efflandt (efflandt_at_xnet.com)
Date: 07/15/03


Date: Tue, 15 Jul 2003 01:32:01 +0000 (UTC)

On Mon, 14 Jul 2003 17:11:23 +0200, Dragan <gekko@eunet.yu> wrote:
> I have a class C internal network divided into 3 sections (one central
> office and two branches) connected by 2 routers (DSL router). Routers are
> communicating throuth RIP2 protocol. Machines are Win98 and WinXP, servers
> are Linux servers (Samba and SQL used), Windows adresses are leased through
> DHCP. The network looks like this:
>
> 192.168.1.0/24 network (around 10 computers) - 192.168.1.1 server,
> 192.168.2.1 router
> |
> 192.168.1.2
> 192.168.0.3
> |
> 192.168.0/24 network (50 computers) - 192.168.0.1 server, 192.168.0.3 and
> 192.168.0.2 routers
> |
> 192.168.0.2
> 192.168.2.1
> |
> 192.168.2.0/24 network (20 computers) 192.168.2.3 server, 192.168.2.1 router
>
> It works fine, but now we want to connect central office (192.168.0.0/24) to
> the internet. We have one public IP adress and we would like to use IP
> Masquerade. I have set up a Red Hat 9 Linux as a NAT server with adress
> 192.168.0.10, IP Masquerade works fine, but now there is a problem with
> internal routed network. I had to set up 192.168.0.10 server as a default
> gateway but that breaks connection with other two subnetworks. If I define
> static routes to two subnetworks on each of the Windows machines then it
> works fine, but I can't set up static routed through DHCP, and I know of no
> other way to define routes other than typing route add... in command prompt.
> If static routes are not defined then everything that goes out of
> 192.168.0.0/24 nework goes to 192.168.0.10 NAT server, where it gets lost.
> If it were only one internal network then it wouldn't be a problem but this
> way I don't know how to solve the problem. Tnx in advance.

If you have a proper router set up, each box on each subnet should only
need a local network route (for its directly connected LAN) and default
route to the local router. Then the local router should be properly
configured to route anything else to the other subnets or default to
internet gateway.

I do not know if you have a central router to coordinate things, or if you
really do have one network chained to another network chained to another
network, and a router in the middle gets confused which way it should send
specific traffic. Perhaps you need a more sophisticated routing protocol
than RIP, or failing that, maybe proper static routes on the routers would
help.

A couple of examples: the Cisco in our office used to be the default route
for our office subnet (and local dialin PPP proxy arp to our LAN), which
would route everything to our factory WAN over frame relay. A Cisco
router there would decide whether to route the traffic locally, to another
office subnet, or to our RedHat masq box to the internet. I think the
Cicso routers were using EGP routing protocol.

Now that changed and we use a SonicWall on sdsl, which is the default
route for our office subnet, and the SonicWall decides what should go
through VPN to our factory WAN, or what should NAT directly to the
internet. Our local Cisco is now configured with a static default route
to the SonicWall. So a dialin PPP has a default route to the Cisco it
connects to, the Cisco has a default route to the SonicWall, and the
SonicWall decides between factory VPN, or NAT to internet. And it is all
done with just local LAN route and default route on local PCs, and static
routes on our Cisco or SonicWall.

But our traffic is primarily between our remote offices/foreign factories,
and our main factory. The remote offices do not really communicate
directly between each other (just to our main factory mail/document
retrieval, and HP3000 computer for order entry/inventory). So a star
topography is all we really need for our factory WAN.

-- 
David Efflandt - All spam ignored  http://www.de-srv.com/
http://www.autox.chicago.il.us/  http://www.berniesfloral.net/
http://cgi-help.virtualave.net/  http://hammer.prohosting.com/~cgi-wiz/


Relevant Pages

  • Re: [SLE] connection redundancy
    ... which is the router just this side of the ... >>because it's own gateway is still up. ... You'll also need three network cards to put into this Linux box. ... your internal LAN can route packets to the Internet. ...
    (SuSE)
  • Re: please advise - problem with routing
    ... and a network is a very important distinction in IP address configuration. ... you show that the 192.168.1.0/24 subnet has two devices ... ROUTE PRINT on NT4 ... ROUTE PRINT on router ...
    (microsoft.public.windows.server.networking)
  • Re: Permanent Branch Office VPN
    ... > Many thanks for your comments - it seems that the 1 x NIC route is the one ... >> MUST use single network card setup on SBS2003. ... >>> router to router VPN connection. ... to the LAN side of the network at head office. ...
    (microsoft.public.windows.server.sbs)
  • Re: 1 server 2 nic 2 adsl lines, gateway problem
    ... A network can only have one default route. ... neither a 'windows' thing, nor an 'SBS' thing, it is pure networking) ... the .1.254 router. ...
    (microsoft.public.windows.server.sbs)
  • Re: Can I Remove This...
    ... Someone else could have a router in the area that's never been configured ... beyond the factory defaults. ... Now that you've changed your unit's SSID, ... may be able to see the other network. ...
    (alt.internet.wireless)