The patch level I'm using is .2188 on all machines.
Running the "ping -I eth1 yahoo.com" results in an unknow host error. All troubleshooting techniques that I would use for normal network problems seem to say that traffic isn't getting past the machine when I have the 2nd NIC running.
The separate IPs are set up on two physically separate NICs so I am not using aliases. Since the 2 different class C's are running over the same switch plane I did initially try to use virtual/aliased IPs on the main eth interface but when I had problems decided to just enable the 2nd NICs.
The newest machine that I set up is using patch level 2115 and I let Fedora do the networking setup via the graphical installation and I gave each ethernet interface an available IP on the two different networks but it still has this same problem. I'm reading that entire HOW-TO now and hope to find something to help. If anyone has anything else for me to try, I'd appreciate any pointers.
Thanks, Kevin
Alexander Dalloz wrote:
Am Mo, den 17.05.2004 schrieb Kevin Kimmell um 15:28:
First: please do not hijack foreign threads. Do not reply to articles you won't reply to but start with an empty new email editor window. Your posting now appears as a reply to "Re: How to install?"
I've got several HP Proliant servers running mostly Fedora versions of redhat. I'm using kernel 2.4.22-1 right now on the ones that are problematic.
Important is the patch level numbering. So 2.4.22-1.2115 is initial FC1 kernel, .2188 is latest bugfixing kernel.
These MBs on the HPs have dual gigabit ethernet controllers in them. /proc/pci reports them correctly as "Broadcom Corporation NetXtreme BCM5704 Gigabit Ethernet (rev 2)".
I've got two public class C networks with two separate ISPs. When I initialize eth0 to either of my public IP ranges with their respective gateways everything works just fine. So now I'm trying to add an IP from the other ISP's range so that I've got multiple pathways to these public servers.
Whether I add a virtual copy of eth0 with the other ISPs IP/gateway combo or if I plug eth1 into the switch and give it one, networking stops working. I should mention that both ISPs come in to the same router and are in the smae switch plane so all wires from the servers are on the same set of switches.
Here's the route table when things are ok: Kernel IP routing table Destination Gateway Genmask Flags Metric Ref Use Iface 204.117.218.0 * 255.255.255.0 U 0 0 0 eth0 169.254.0.0 * 255.255.0.0 U 0 0 0 eth0 127.0.0.0 * 255.0.0.0 U 0 0 0 lo default 204.117.218.254 0.0.0.0 UG 0 0 0 eth0
What the hell is that 169.254.0.0 statement? I've searched hi and lo and can't figure it out. I'm making a blind guess that it's an IPv6 thing? I didn't put it there and removing it doesn't survive a reboot.
That was now explained several times here on the list. See http://www.zeroconf.org/ and the list's archive. It has nothing to do with IPv6. How to deactivate it please consult the lists's archive.
Anyhow, at this point all networking is fine. I can ping and trace out and vice versa. Now, when I initialize eth1 here's the table:
Destination Gateway Genmask Flags Metric Ref Use Iface
12.168.88.0 * 255.255.255.0 U 0 0 0 eth1
204.117.218.0 * 255.255.255.0 U 0 0 0 eth0
169.254.0.0 * 255.255.0.0 U 0 0 0 eth1
127.0.0.0 * 255.0.0.0 U 0 0 0 lo
default 204.117.218.254 0.0.0.0 UG 0 0 0 eth0
Well, the default route is still on the eth0 device.
As soon as this is activated, networking stops functioning. I can't ping in or out. Am I missing some key step? I know that it works becasue I'm running a firewall on another machine that's running a 2.4.20 kernel and it was autoconfigured and it's functioning properly.
I would imagine if you run "ping -I eth1 $target" it will result in ping replies?
Can anyone give me a clue?
Check
http://lartc.org/
http://www.tldp.org/HOWTO/Adv-Routing-HOWTO/
for advanced routing howtos. You will have to use advanced features to handle both different uplinks / class C nets.
Thanks,
Kevin
Btw. how did you set up those 508 IPs for this server? Do you use aliased ethernet devices or the more modern setup way using iproute (command ip)?
Alexander
