error on ethernet interface

skoinga · 11-02-2010, 08:28 AM

Hi,

I've a server with two NIC.
Sometimes the server becomes unreacheable. When this happens, I did a reboot and then it works.. but when looking at the interface status (with ifconfig command), I've notice several "errors".
So, I've tried to change NIC.. but the problem occurs again.
I've also changed the port on the switch.

Any suggestion? Could be a software problem?

stress_junkie · 11-02-2010, 08:37 AM

Are both NICs connected to the same router? If so do they have different names and software addresses?

I believe that if they are connected to the same router then they should have different names and software addresses. Otherwise the router may become confused when mapping the hardware address to the software address.

But I could be wrong.

kirukan · 11-02-2010, 08:40 AM

If it is linux based, check the following,
lspci
dmesg
route -n
/var/log/message
less /etc/sysconfig/network-scripting/ifcfg-eth0 and ifcfg-eth1

mesiol · 11-02-2010, 08:48 AM

Hi,

also check cabling and if available logs on your switch/router.

Skaperen · 11-02-2010, 01:27 PM

Quote:

Originally Posted by stress_junkie

Are both NICs connected to the same router? If so do they have different names and software addresses?

I believe that if they are connected to the same router then they should have different names and software addresses. Otherwise the router may become confused when mapping the hardware address to the software address.

But I could be wrong.

I have many machines with multiple NICs configured with the same address. This is for fallback redundancy, not load balancing. To make this work right, I set rp_filter to 0 to disable RFC3704 Strict Reverse Path and RFC3704 Loose Reverse Path (intended to prevent in-LAN DDos attacks via IP spoofing). This allows asymmetric paths. If you trust your LAN, this should be OK.

Code:

(for f in /proc/sys/net/ipv4/conf/*/rp_filter;do echo 0 >$f;done)

Collect the IPv6 link-local addresses of your machines with 2 or more NICs. On another machine, ping or SSH in to the machines with non-working IP addresses to see if these IPv6 addresses are working. On the client, you will need to append "%eth0" to the address (or replace "eth0" with whatever client interface can reach the server target NIC). Even if the IPv4 addresses, ARP, or routing gets messed up, if it is on the same LAN with a transparent switch, IPv6 should be able to connect. Try the connection against each interface. If you can get in via SSH, it might be more convenient to diagnose trouble since you are at another machine that can be try to make connections to IPv4 addresses.

Code:

ping6 fe80::215:17ff:fef2:63e8%eth0
ping6 fe80::215:17ff:fef2:63e9%eth0
ssh root@fe80::215:17ff:fef2:63e8%eth0
ssh root@fe80::215:17ff:fef2:63e9%eth0

szboardstretcher · 11-02-2010, 01:53 PM

check the cabling