[Dnsmasq-discuss] Set outgoing IP-Address to dns-upstream-servers
dnsmasq.list at michaelrack.de
Fri Mar 6 15:51:53 GMT 2009
Well, in normal the routing have to be correct.
But my network is very complex. I have three upstream carriers with one
static ip-address for each connection. Every interface is different in
connection-type and connection-speed.
Over all interfaces i run a VPN-Tunnel 3x /30 Networks with public
ip-addresses to interconnect with a Root-Server in Nürnberg (Germany).
A dynamic routing-protocoll set up my routes automaticly, even the
This setup allows me, to have a allways reliable innternet-connection.
My local-router in Salzburg (Austria) announces a /28 public ip network
to serve all services behind the router.
Now to the Network-Setup:
tap1: 91.xx.15.49/30 < 91.xx.15.50/30
tap2: 91.xx.15.53/30 - 91.xx.15.54/30
tap3: 91.xx.15.57/30 > 91.xx.15.58/30
locally assigned: 91.xx.15.113
< and > indicates the way the packets will be sent (asynchron routing)
- indicates a backup-link if one of tap1 or tap3 will fail
Now, i like that DNSMASQ use 91.xx.15.113 for querys to upstream
My dynamic routing-protocoll inserts this routing-entry into the
default via 91.xx.15.49 metric 1
no src-option is given by the dynamic routing-protocoll. The right way
for all reasons is to define src 91.xx.15.113 if the packet does not
contain an ip_src.
But my routing-protocoll doesn't support this.
Now DNSMASQ have to set the ip_src address, to supress linux-kernel to
complete the outgoing packet.
Am 06.03.2009 16:20, schrieb Rance Hall:
> You do this with routes, on a multihomed machine each IP address is
> also part of a network and is part of the route to get to that
> If you set up your routes correctly (generally speaking the default
> is correct) then the correct local static ip is always used when
> leaving your box to go anywhere, including upstream dns servers.
More information about the Dnsmasq-discuss