Troubleshooting NIC Issues

[tobyheywood]

Hi,

I have a SuperMicro 1U server with 2x Broadcom 1000MB NICs. I'm using Slackware 10.1 and kernel 2.6.8.11, with BCM5700 (version 8.1.3) patched into the kernel.

I have recently begun to experience connectivity issues to the server, which happen a few times a day. However the pattern which seems to be emerging is that I lose connectivity when I am in the middle of a ssh session and I have begun to do some which outputs a lot of text to the ssh console.

For example, i ran

rm -fvr pureftpd/

And then suddenly found that the server had become uncontactable.

Could anyone offer suggestions as to the best way to diagnose this problem and also if anyone has had a similar problem it would be great to hear about your experiences.

Toby Heywood

http://www.tobyheywood.co.uk/

 

[TrojanWarBlade]

Are you only having problems with ssh or with other connectivity as well?

BTW: Nice to see another biker here! ;-)


Trojan.

 

[tobyheywood]

Trojan,

Everything is fine up until the unknown problem happens, once the problem has appeared all connectivity via any eth dev is lost.

There are absolutely no error messages being produced, which makes me wonder if it is just the NIC, I would have thought that if an error occured that something would at least be logged.

If the server is locking up, then this is quite concerning, especially when considering that during testing none of these problems were showing and now that it has been promoted to a production machine it just makes matters worse.

I'm thinking of just recompiling the latest kernel with the latest and greatest Broadcom drivers patched in, although it would be nice to find out what is causing this problem.

Are there any specific things I can do in order to get the network card to basically provide debug information. I'm currently reading up on ethtool to see if that will be able to shed some light on this rather troublesome issue.

Also any suggestions as to the best way of pin pointing it to the NIC or the server would be good.

On a seperate note, Trojan what are you currently riding?

Toby Heywood

http://www.tobyheywood.co.uk/

 

[TrojanWarBlade]

Do you have a spare NIC you can swap this one out with?
I doubt that ethtool will help much, you'll probably see normal transactions and then nothing.
My first thought would be to try the hardware swap.
Also, do you have a completely different card to try if the first swap fails? It might sound crazy but it would save you a risky recompile since the new card would probably use a different driver.

On that different note, no I am not riding at the moment.
I haven't been working for some time now and I couldn't afford to keep up the repayments so I had to let my beloved Suzi go.
:-(


Trojan.

 

[tobyheywood]

Unfortunately the card is of the 'on board' variety so will be difficult to perform a straight hardware swap. I think I could probably get my hands on a replacement NIC.

Any recommendations on make, in the past I've used Intel/3Com NICs. Any you would recommend to stay clear of?

TDN: Shame about the like of wheels. :-(

Toby Heywood

http://www.tobyheywood.co.uk/

 

[TrojanWarBlade]

Not really got any good suggestions for cards to try but I do think that adding a new card would be worth a go. It's quick, simple and relatively cheap and could solve your problem instantly.
Worth a try in my books.



Trojan.