I have an odd problem on a server at work the network interface stops responding for no apparent reason..

A couple days after being setup I came in to work, and when I checked my email, Thunderbird timed out contacting the server.. I tried to hit the webmail interface, and that timed out as well.

I just setup the server last week as a mail server running Debian Stable (loaded with 3.1r4) running postfix, courier, and not much of anything else at the moment. The old mail server died a harsh death. It was a PIII 550 with 128 MB of RAM which was a temporary fix 3 years ago when the last Windows mail server died, I've been asking for new hardware ever since then, typical response was... "mail is working request denied..." Well the request was finally granted after that server died, and I have a new server on the way but I need this temporary one to work until I get the new server loaded just the way I want it, before putting it into production.

I went to the server console and did a ifconfig -a and everything looked normal.

ifdown eth0 gave me some error which I don't recall at the moment, possibly something about not defined.

ifup eth0 returned no response, but after which the server, and network interface worked fine again...

I looked in the logs but didn't see anything that looked out of place. so I just kind of wrote it off as a fluke. well apparently it's not a fluke because I just tried to reach that server and it's down again.. and since nothing showed up in the logs as a blatant error last time I thought someone here might be able to toss me a clue of someplace else I could look to determine what is going on..

I believe the system is a Supermicro X5DAE Motherboard with Dual Xeon 2.8 Ghx CPU's and 4 GB of memory.. 3Ware 8006-2 SATA RAID Controller with DUAL Seagate SATA Drives.

This was the only hardware I had available at the moment as a temporary fix (much better than my options last time around).. I just can't understand why the Network Interface on the Motherboard is dying a silent death..

Thanks for any pointers you can offer.

I would say this is a hardware problem if you are not getting errors recorded in the log file. I had a similar problem over a few months. I replaced the offending ethernet card and have had no more problems.

Perhaps the network is shut down automatically after a period of inactivity? What does the BIOS tell you about it's setup?

Well considering this hardware was a Windows server that was in operation for over two years and I never had an issue with the network shutting down I don't think theres any setting in the BIOS that would cause the network to stop functioning after lack of use.. and considering it's a email server constantly being hit by spam I doubt theres much idle time on the interface. .

I should be able to throw a new network card in it on Tuesday..

I certainly won't rule out a hardware problem as this machine did do a very bizarre thing right before I removed it from service as a Windows server.. The system 'forgot' that the admin accounts (both local and domain) had admin rights. I couldn't install anything new, or make any changes to the system that required admin rights. and I have never witnessed that type of problem before.. I never did figure out what caused that problem either..

I was wondering if you guys had any insight on my issue. On my debian box, I'm running 5 as well but with 2 network cards. One network card has an internal IP address and the other is external. It's a LAMP setup on the machine but I get this weird problem. On the external IP addresses, after a period, it stops responding to any http requests. I tried to restart apache with /etc/init.d/apache2 restart but it doesn't do anything. I can still SSH into the box with the internal ip address and the website still responds with the internal IP address but it just doesn't.

I don't know if it makes a different but I'm running something called ISPConfig 2 (http://www.ispconfig.org/), which is a LAMP type software. The only way I can get it back up is to soft reboot the machine.

Do you guys think it might be a faulty network card?

Could be dodgy network card. When it stops working the restart networking because it isn't Apache but packets not going through;

# /etc/init.d/networking restart

you can observe status with # netstat

Rob

Thanks so much for the reply Rob. I'll check it the next time it just craps out and report back!

It just crapped out and I restarted the network and the site was back online, thanks for the tip. I think the network card is busted and has to be replaced.

If I insert a new PCI card, will Debian lenny 5, automatically pick it up via PnP(ie. eth3, whatever the next one up is)? I've read that MOST network chipsets are supported with Linux by default. I just want to get one that works for sure without any hassles with adding modules to the kernel and all that stuff (which I don't know how to do).

Strangely most of the really cheap cards work fine. Have a look at the specs for the card as most manufacturers will actually state what operating systems are supported.

Rob

Besides the hardware issue, will not hurt checking your routing table, i'm having an issue with a solaris 10 and 11 box running zones, with two IP addresses on different sub-nets (1 address for development server and 1 address for QA server) , and the default routing algorithm is round-robin, so i can ping to sub-net A for a while, then it dies and can only ping B sub-net, but only one is active at a time. Like i said, wont hurt and prolly is the network card