multiple ips, multiple gateways, one interface

drpixel · 11-25-2002, 06:15 PM

Greetings all!

I'm not much of a TCP/IP person so please forgive me if this is a simple answer, but I think I've tried everything I can think of. Here's the problem. I run a small co-located server that I currently host my friends on, as well as a few clients. I do IP based hosting, and have just run out of my initial allotment of IPs from my ISP.

The two blocks of IP's are on different subnets, and I'm having a tough time getting it configured. Here's an example of my IP blocks. (IP Addys withheld to protect the guilty!)

I have included a lot of detail so that someone smarter than me can double check my math!

Block #1
-----------
network: xxx.xxx.177.64
gw: xxx.xxx.177.65
range: xxx.xxx.177.66 thru 78
netmask: 255.255.255.240

Block #2
-----------
network: xxx.xxx.144.48
gw: xxx.xxx.144.49
range: xxx.xxx.144.50 thru 61
netmask: 255.255.255.240

The server's primary IP address is xxx.xxx.177.66, and default gateway is xxx.xxx.177.65. All aliased IP addresses work on the first block, no problem. I attempted to add an IP alias on the second block by first using 'ifconfig' like so:
ifconfig eth0:1 netmask 255.255.255.240 broadcast xxx.xxx.144.62 xxx.xxx.144.51

then I attempted to add another default gateway, using:
route add default gw xxx.xxx.144.49 dev eth0:1

I can ping myself with no problem when logged into the box via SSH, but from remote, I get the dreaded 'Destination Unreachable' errors. I know I'm not doing this right, can someone please tell me what I'm not doing, or doing wrong, or at least which bridge to jump off of?

Thanks!

rioguia · 11-26-2002, 02:35 PM

what is the output from:
route -n

drpixel · 11-26-2002, 03:03 PM

Here's an interesting side note... I just rebooted the server, which of course cleared the ifconfig eth0:1 step mentioned earlier. When I attempted to add the alias interface back using the procedure listed above, I got:

Code:

[root@cp root]# ifconfig eth0:1 netmask 255.255.255.240 broadcast xxx.xxx.144.62 xxx.xxx.144.51
SIOCSIFNETMASK: Cannot assign requested address
SIOCSIFBRDADDR: Cannot assign requested address
SIOCSIFFLAGS: Cannot assign requested address

Now running ifconfig again shows that the interface is UP and RUNNING. is this abnormal?

Here's the output of the route -n command you requested.

Code:

[root@cp root]# route -n
Kernel IP routing table
Destination     Gateway         Genmask         Flags Metric Ref    Use Iface
xxx.xxx.144.48  0.0.0.0         255.255.255.240 U     0      0        0 eth0
xxx.xxx.177.64  0.0.0.0         255.255.255.240 U     0      0        0 eth0
127.0.0.0       0.0.0.0         255.0.0.0       U     0      0        0 lo
0.0.0.0         xxx.xxx.144.49  0.0.0.0         UG    0      0        0 eth0
0.0.0.0         xxx.xxx.177.65  0.0.0.0         UG    0      0        0 eth0
[root@cp root]#

rioguia · 12-03-2002, 08:48 AM

Sorry for the delay in responding (birth of a second child has really slowed me down). I'm not a routing guru. I had a similar (but different) problem trying to route do both a bridge and a router on a linux box. see
http://www.linuxquestions.org/questi...threadid=31882

I would guess that your problem is that your outgoing packets (the zeros in the two left column entries) have two possible addresses (144.49 and 177.65).

Quote:

0.0.0.0 xxx.xxx.144.49 0.0.0.0 UG 0 0 0 eth0
0.0.0.0 xxx.xxx.177.65 0.0.0.0 UG 0 0 0 eth0

Its unclear to me how the packet would know which address to pick. I'm sorry but I don't know the solution to this but the moderator Peter Robb usually has the solution (i'll be happy to look at it the next time i'm on a linux box).

drpixel · 12-03-2002, 03:15 PM

I appreciate the help! I too was wondering how it would magically know which gateway to use, other than to track which IP the packets came from, then reply through the appropriate gateway.

Seems like there has to be a way to do this somehow. I'd think a lot of ISPs and webhosting companies would run across this problem quite often as the expand their business. (although I'm sure careful planning would help eliminate this a lot!)

rioguia · 12-03-2002, 03:58 PM

if i had to guess, i'd say that you need a bridge to do what you want to do and you haven't mentioned anything about that. all of the ip aliasing address how-to's I have seen, have used examples of ip addresses all on the same network:
see e.g.:
http://www.tldp.org/HOWTO/mini/IP-Alias/commands.html

for information on a bridge, try:
http://www.linux.org/docs/ldp/howto/BRIDGE-STP-HOWTO/

if you think I'm wrong, could you give the actual output of

/sbin/ifconfig -a (redact IP addresses if you like)
/proc/net/aliases
/proc/net/alias_types

clavius · 12-04-2002, 12:56 AM

Quote:

Originally posted by rioguia

if you think I'm wrong, could you give the actual output of

/sbin/ifconfig -a (redact IP addresses if you like)
/proc/net/aliases
/proc/net/alias_types

/proc/net/aliases ? which kernel version ?

You cannot expect to have 2 default routes with 2 different gateways unless you are using dynamic routing and assign to them different metrics. That's because only one can be used at a moment time. And the default used is that which is first in the kernel routing table, unless you are using some metrics.

Be carefull... linux marks packets with eth0's ip source even the gateway is in the network of eth0:1 IP ;( This behavoir I've discovered recently...