Network connectivity flaky, fine on other systems

[exodist]

Interface comes up fine, gets ip and router. I try to use the interface and it works fine, can go to google etc. However after a minute or 2 of usage it will suddenly stop, I will try to ping and ping will sit there silently, after a varying ammount of time it will start to work again, when it does the browser goes around fine and ping starts giving me messages, if I have ping going when it suddenly starts again it will give me giant numbers for ms timing.

I have a gigabit network going here, I have 5 computers connected to it and all but 1 are fine.

Basically this computer has 3 nic's, one for a direct gigabit (crossover cable) connection to an nfs server, one is a connection to the gigabit network all my systems use, then the last one is to a dsl router that provides a static ip. 3 systems have this exact same setup. I used the multipath routing to support this, here is the tutorial I wrote on it: http://wiki.linuxquestions.org/wiki/...load_balancing

2 of the systems work perfectly fine, the third does not. I am pretty sure it is not my dual ethernet script that is causing this, just to be sure I tested by starting w/o the dual ethernet and using only one interface configured via dhcp and the problem still occured.

I have tried reconfiguring the kernel with every option I can think of for why this is happening.

I have tried several network cards and all have the same problem:
usb 10/100 (pegasus)
3com 10/100 (3c905btx)
linksys gigabit (realtek)
onboard (nforce)

Strangly enough however this does not happen on the direct connection tot he nfs server, that connection is rock solid.

I have tried changing what port the computer is connected to on the switch, and I know the switch is good because I have other comps connected to it not experiencing problems. When the internet is in flake mode on this system others have no problems.

Kernel 2.6.18, I have tried 17 and 19 as well
Gentoo linux x86
intel core 2 duo e6600
evga nforce i680 chipset
1gb ram

--Added:
I have checked and verified this problem occurs on the dsl interface as well, it seems any interface used for internet connectivity has the problem, but ones w/o internet connectivity do not.

[exodist]

I ran another set of tests, 2 systems, both 64 bit, one amd64 on an old asus board, the other e6600 on the evga i680.

The test:
all of the following simeltaniously
download a dvd iso of debian linux from an http mirror
endless loop of downloading the google frontpage (while [ 0 -eq 0 ]; do wget www.google.com; done)
and ping 192.168.0.1 (my router)

on both systems this process ran for 15 minutes or more

the i680 system using a pci gigabit card had 55% packet loss, and the debian iso failed it's download
the i680 system using a usb 10/100 card had 37% packet loss, and the debian iso also failed it's download on this one.
the amd64 system when it was done I did a control+c to stop the ping, 0% packet loss

same router, same kernel version and basic config (slightly modified for different cpu's etc.)
Go to Top of Page

[exodist]

I am an idiot and must appologise. This is indeed a problem on my end... I tried everything I could think of for 2 weeks before making a report because I wanted to be certain it was not my end before wasting anyones time...

The problem was that I had a box hidden away under my desk that I had forgotten about when I made my new ip scheme. There was an ip address conflict, I resloved it and there are no more problems, and I can even justify the test results showing eth0 as good and others as bad:

The tests where they were initilised as eth1 I used the network scripts that gave them the static ip's, this was conflicted, obvious bad
the tests where it was eth0 I used dhcpcd real quick, thus no ip conflict.

I am at a loss of words for anyone who's time I have wasted w/ this.