network goes down

[jakess]

we using TCP/IP and NT4 and Win95 with CAT4 cabling and a few switches together with a 10base non-intellegent hub.
every 2 weeks the network starts giving problems like users can't log on or lose their network connections or their network applications becomes very slow and causes their pc's to hang.we had a sniffer here before and traced a few faulty NIC's that was replaced.we also replaced some network cables.this problem is however still carrying on and is our only solution to reset all our switches and hub...power on and power off.
this is just a quick fix and would i like to hear some of your comments please....

 

[rfeenie]

odds on it's pointing to your CAT4 cabling, may need to move over to cAT5e or 6 if its financially feasible. If in doubt get the Ball Pein Hammer out

 

[wybnormal]

I doubt its the cabling since it runs fine for 2 weeks or so. I would look more at things like the NT server having a memory leak and screwing up the DNS/WINS ( seen this first hand), check the RX Buffers on the switches for high error counts ( 3com infamous for this) the switch would slowly go "stupid", try to reset only a single switch at a time to isolate, possible spanning tree issue( a real stretch honestly).

My first step the next time is to reboot the DNS/WINS box, then PDC. Next the switch the the servers are tied to and just work your way down the chain.

Mike S

 

[mbhandari]

I agree with Mike , I feel most likly this problem be with switch as thay tend to Buffer problems This some times happen with Baynetwoks Switches also.
-bhandari

 

[jakess]

a lot people said i should look for a faulty NIC or cable.
we are using cabletron swithes and have we replaced some of it already. i am also of the opion that is a high collision rate on switch is causing it, but what is causing it as we have swopped them around. we have aslo unplugged all the network cables and then plug them back in one at a time to test the lines but learnt nothing from that.we will be upgrading to CAT5 and fibre-optic this year but i need to give some explanation why is is happening.
Mike S you talking about a memory leak on a server messing up the DNS/WINS....how should i troubleshoot this??
our PDC is also the DHCP/WINS server.

 

[wybnormal]

The memory leak can be seen over time( most of the times) by keeping either perfmon running with the memory logs enabled or if it's bad enough, just using task manager with it fliped to memory usage. You will see the committed memory slowly crawl up and not come back down. Personally, I would split the DNS/WINS/DHCP from the PDC. That is MS's own suggestion and I have seen it cause problems in the past on my own networks with it all running on the same box. The great thing about DNS/WINS/DHCP is it can be a small cheap little box thats too "small" to be a "server" but it would work great for this.

You said you think a high collision rate is causing this? why do you say that? lights? stats? If you have swapped the switches and it's still there, it's a crimped cable which would show up all the time OR a workstation is failing or has a badly behaving application on it. Check your connectors. I have been surprised by a "failure" that was really a RJ45 that had not been plugged in all the way to click. Darn boots.. A lesser thing but I have seen it is where the cables have been run to close to the overhead lights or power runs. The emitted RF noise from floresence lights and house power lines can wreak havoc on your network and it show up as collisions or a high error count. I moved a cable 2 feet one time and "fixed" a long standing problem of errors.

Mike S