how to find a bad network device?

[phreaq]

I have a bit of a problem at work with our network, and I'm hoping for some help

One of our SQL applications loses the network connection every so often, and results in some data loss. I've done some inital testing and I believe there is a fault of some kind in the actual network cabling (the problem spans different computers at different times). It could either be a bad cable, router, or card, or bandwidth, or whatever.

I'm wondering if there is some kind of software or hardware test I can run on our network all day long and see how often it 'drops' out. Anyone have any thoughts?

As for our network, it's got about 8 servers, 80+ computers, ethernet running thru 10/100 switches or 10 hubs.

any help would be greatly appreciated

phreaq
Has anyone seen my brain today? (^_^)

http://www.audiopornproductions.com

 

[robertjo24]

Check the error logs in the switches (if they are managed switches) for ports receiving errors. A faulty/intermittant cable will log errors on a port.
Check your router interface statistics for errors. Same as above.
Is there anything in common withthe nodes experincing issues, (i.e. same applications running at time of outage, outage at high utilization periods, massive SQL queries at the time of loss)? If you have a specific application that is causing errors, you could sometimes cause port errors with a faulty application. Are all of your recommended patches in place for each application you are running?
You mentioned hubs, do any of these machines connect to hubs, then to one switch port on a switch? If so, you could be running into collision issues causing a switch port to drop. This is especially nagging when there are several computers on a hub that is plugged into a auto sensing switch. The switch could be dropping the port due to errors.

There are several network monitoring tools available on the WEB.
We used to use WhatsUp Gold. It is a fairly powerful/versatile tool. It will page or email you when a node drops.

 

[DTSMAN]

I am have the exact same issue. Our app is running on Sybase SQL. We replaced the hub and the cable without change. The server is a Dell with XPpro on it and this is the only site having the issue. Of course we have to exhaust all possibilities of the software before we tell him it is his Dell. It seems to be an instant hiccup and the error logs aren't picking it up.

Bo

I'm a man,, I can change,, If I have too,, I guess. (Red Green)

http://www.redgreen.com/

 

[Sympology]

Other possibilty we can across regulary, NIC settings.
If you switch is set say to Auto Negotiate but the server is set to say 100/Full duplex, you get no end of problems.
Make sure BOTH ends are set the same be they 100/Full or Auto Neg (or god forbid 10/half).
A good indicator of this is if the switch is set to 100mbs set what the duplex is, if it's half, it's almost certainly a mismatch.


Only the truly stupid believe they know everything.
Stu.. 2004

 

[jimbopalmer]

Stu weirdly, 10/half gives much greater throughput than 100/full mismatched to Auto-negotiate, as the auto (N-Way) protocol defaults to correct half duplex behavior if the other side does not also auto negotiate.
100/full mismatched to auto can give under 1 megabit of throughput.

Never set full duplex on any wire you cannot control on BOTH ends.

I tried to remain child-like, all I acheived was childish.