Any help would be great - Major network problems

[jumpinjamez]

Hello,

 

I have major network errors and like any advice i can get at the moment.

 

network history

 

550 - Worsstations

 

12 servers

 

20 netgear switches

 

5 fibre connection with convertor to copper

 

 

Last Wednesday some smart student decided to loop back the network by plugging in a network cable in two port in a room.

 

this problem was found and fixed quickly however since that day the network has just gone crazy.

 

 

It is very very slow loading profiles - when it works.

 

At first it was thought it was the profile server so all roaming profiles were moved to another server however now when the user runs office it hangs on the configuring office for the first time.

 

Switches are failing then working again.

 

It all points to network hardware however it seems so random. It works somtimes then it does not

 

Server event logs dont say much that links to this.

 

DHCP , DNS allworking OK.

 

Any ideas of areas to look at would be great.

[petectid]

It will be helpful if you posted the type/model of the Netgear switches.

[witch]

This happened to us (on a smaller scale) and was only fixed by switching everything off and then slowly putting it all back on bit by bit

[marco84]

I would start by re-booting each one of your switches.

[Guest]

Would try powering down whole network and the bring up your switches, then servers, then workstations etc

 

Also seeing you have netgear kit, make sure they are all running the most uptodate firmware. HAve had netgear switches in the past that have shipped with well out of date firmware that has caused problems.

[ICTNUT]

I recommend what WITCH says above, switch everything off, bring the switches backup one be one, no depending on your setup you may need to do these in a set order especially if they are chained.

 

Next bring the servers back up, but only the essential ones that let you logon to the network, the startup a client pc and see if you can logon and monitor speed.

 

If this is ok start bringing up other servers and while doing this start and restart the same test client pc to see if there are any issues highlighted.

 

If all of this is ok, move around the school and start client pcs in different locations to see if there is a problem, you will soon know if there is.

 

At the end of the day if you network is still running this slow you could have a second or third loopback somewhere.

 

Once you have sorted the problem the next step would be to see if you switches have any form of storm control and then enable it.

[techyphil]

We have also had this problem. The network went crazy and again I went around and found the kid has looped it around.

 

Gave us a great opportuity though to say some of the kit bust and got funding for newer better kit.

 

The kit replaced though was very old and so overall good job the scare happened else we wouldnt be running 1GB links everywhere.

 

Yeh good suggestions btw, reboot the lot mate

[Geoff]

Also see if you can enable STP on your switches this should hopefully prevent loops bringing the entire network down, or at least contain the problem to one or two switches.

[StewartKnight]

I was just about to say Spanning Tree Protocol too! I'd recommend managed switches and turning STP off!

[DMcCoy]

Why would you want to turn STP (RSTP) off?

[projector1]

Why would you want to turn STP (RSTP) off?

 

if the loopbacking is done regularly by the students then yes enable it, it does have a overhead though!

[TechMonkey]

I was told that STP should be on between switches but off for stations. If it is on for stations it can delay the network negotiation up to as much as the length of time it takes for a PC to be booted and logged on (probably an older machine but not a relic), thereby missing policies. Between switches it at least means that the loopback is isolated to that segment and the rest of the network will carry on regardless.

[DMcCoy]

Why would you want to turn STP (RSTP) off?

 

if the loopbacking is done regularly by the students then yes enable it, it does have a overhead though!

 

Even irregularly is enough for me to have stp enabled for all my managed switches!

[Geoff]

That advice was correct about 5 years ago TechMonkey. The successors to the original STP protocol (802.1D and 802.1Q) correct these deficiencies.

[jumpinjamez]

Thanks for all the support.

 

After the big Power cycle (turned it all off and back on again)

 

tracked it down to a faulty port on the main network switch great network all back up.......however as the profile server keeps rebooting all students profiles have been moved to another server and when users first run any office application the computer hangs and does nothing else at Prepairing to install Office screen.

 

Maybe a corupt default profile i think i am checking now.

 

 

Again thanks the help, as you can guess with the whole school network down i have had lots of Fun

[Kyle]

Also see if you can enable STP on your switches this should hopefully prevent loops bringing the entire network down, or at least contain the problem to one or two switches.

 

Any chance of an idiots guide to doing this as i know nothing about configuring switches.

[spc-rocket]

Spanning Tree is good for preventing loops but you do need to configure it properly. For example it sould only be configured for ports that connects to other switches or routers, it should not be enabled on ports that are connected to servers or workstations.

 

For connectivity problems such as GPO not assigning properly and scripts not running could be because the port that the workstations is connecting is still checking for loops. The thing to is like i said turn of spanning tree for ports that connects to workstations. In cisco switches its called portfast which basically tells the switch to immediatly move to forwarding state rather than checking for loops.

 

Now you may ask what if someone connects a patch leads to two network data points in this case the loop would occur but the switch will not let this propergate it to other switches because the port that connects to other switch(s) will have STP enabled on it.

 

For multiple vlan scenario, rapid spanning tree might be ideal however you do need to calculate the root switch, and the path cost etc. to make it work correctly.

 

Loops can also occur if spanning tree is turned on espically when configuring port trunking or link aggregation or etherchannel as cisco calls it. However if configured correctly the etherchannel is seen as one link to STP instance.

 

Ash.