Network getting too large?

[imperium2335]

Hi,

 

We have had our network setup for a few months and was all working well.

 

But now for no apparent reason, all workstations lose connection to the server and all the shared resources etc for about 10 seconds every 5 to 10 minutes but can be less frequent, it seems random..

 

For some strange reason, connection to the internet is also lost at the same time.

 

I have ruled out the 2 switches on the network as I am able to continually ping workstations throughout the building with no disruption.

 

We have a switch on the ground floor which is connected to the switch in the top floor via a cat6 cable on a 1gbps port. The server is then connected to the switch on the ground floor to the other 1gbps port, all other workstations are connected to the other 100mbps ports.

 

Both switches are Netgear Prosafe 48port switches.

 

The router is connected to one of the 100mbps ports.

 

We also have a phone server that is attached to one of the 100mbps ports (which I can ping continuously fine).

 

We have about 40 PCs, 40 ip phones and 8 printers on the network.

 

Any help is much appreciated.

[AngryTechnician]

Are the "shared resources" shared from the server? Making the server the one thing that becomes unavailable?

[imperium2335]

Hi Angry,

 

Yes, the server contains files that are shared to users. If I constantly ping, I see that every 5 to 10 mins, I get no response from the server address, can't access the net either, which is why I think it's some kind of congestion?

 

Someone suggest VLANs might be a solution but I don't see how that could apply here since the network is in one building with everyone working together.

[Mr.Ben]

More likely the server in my opinion - I have seen a similar thing when one of my network cards started to fail.

 

I don't think your network is too big, I have a flat LAN to support the curriculum network, 120 machines, 40 IP Phones and 8 (virtual) servers

[imperium2335]

But why would the internet fail? Since the server and router are not connected to each other, they are both on separate ports on the switch

 

BTW, each workstation is configured with the server as primary DNS, and the router as secondary DNS.

Edited May 31, 2011 by imperium2335

[maniac]

Check the really obvious stuff first, as you should have absolutely no problems with that size network on whatever switches you're running it on - we have a flat network with 600+ devices connected and get very few problems with it.

 

You say you've 'ruled out' the switches, but if I'm honest I'm 99% sure this problem will be to do with the switch, or the servers connection to it.

 

I would check/do the following.

 

Patch cables between switch and servers

Check firmware on the switchs is up-to-date

Check you haven't got excessive traffic on the network, something constantly sending out packets or a loopback - this can cause weird things to happen.

Maybe force the server to use GB instead of auto negotiating

 

That would be a good start.

[imperium2335]

What's a good way to check for excessive packets or loop backs?

[maniac]

What's a good way to check for excessive packets or loop backs?

 

I'm not familiar with those switches, but if they are managed switches with an interface you can log into, there should be a way to look at the traffic levels on the switch. If it's very high at or near capacity, then you may have a loop. To be honest, it probably isn't the case, but I would check as a matter of course anyway. If you don't have managed switches then just look at the traffic indicator lights on the switch, if they are constantly on even if the network isn't being used much, then you've got excessive traffic.

[imperium2335]

Thanks, I will have a look tomorrow and hopefully all is well.

[bodminman]

Any chance you could get hold of an alternative switch to replace to core one for a day?

[imperium2335]

Unfortunately not We would have to order another one in only if I can confirm this one is broken.

[bodminman]

Where are you based?

 

I'm sure if there is someone on here close by they maybe able to lend a hand.

[imperium2335]

Staffordshire

[bodminman]

I'm in Sunny Suffolk so can't help you with a bit of kit from here,sorry!

 

If the issue is ever present, can't you borrow a switch from another cab and try it out of hours?

[imperium2335]

Could it be caused by one of the workstations? perhaps a virus or something that is constantly sending out loads of traffic? But I can't see how a client on a 100mbps would saturate 1gbps :/

[bodminman]

How many connections to cabs do you have coming into the switch?

[imperium2335]

We only have one Because we are a small/medium business at the moment and not heavily web based just yet.

[steve]

A couple of other things to try -

 

When experiencing issues, pull the plug on the upstairs switch - does this return the workstations downstairs to normal speed. This would narrow down issues to upstairs.

Again when issues are happening, check you can ping the router. It may be this is rebooting (it could affect everyone, depending how DNS is setup on the server)

Check on a workstation with problems if you can ping the server both by IP and dns. Try pinging other resources. Try accessing other shares too.

[imperium2335]

The switch in the cab has 48 10/100 ports, about half are taken up.

[bodminman]

Disconnecting the upstairs switch when the issue is happening may help to isolate its location.

 

Is there anything in the server event logs?

[imperium2335]

Nothing weird in the server event logs, except a lot of XPS print failures from my workstation under 'System' logs

[TheMinister]

You may find that you're switches are being overloaded with ARP traffic, which will force the switch to clear the cache giving the symptoms you've described.

[imperium2335]

Hi,

 

How do I reduce or eliminate this ARP traffic?

[TheMinister]

You can't eliminate it. The ARP cache is like the 'directory' of what's connected to the switch and any other switch.

 

You can reduce it by creating VLAN's or buying a higher grade switch with a larger ARP table.

[imperium2335]

Thanks, I'll have another stab at VLANs. Tried it yesterday but couldn't get the printers to work properly. Shared resources like printers, server and router all need to be in every group am I right in thinking?