Cisco topology change problem

[mjb22]

Hi all,

I am running a 100MB fibre LAN extension to one of my remote sites. Ever since it has been installed it hasn't been working correctly. At the remote site the link is going into a Cisco 3550, at the core a 2950.

From what I can see everytime there is a topology change on the network, it's causing this LAN extension site to become unusable. To get the site backup I have to manually shut down, and bring back up the port it connects to on the 2950. That seems to cure the problem.

I am running EIGRP btw.

Any ideas would be greatfully accepted.

Cheers

Matt

 

[l1carter]

Matt,

at the end you said you are running eigrp??? is this a layer-2 trunking interface or are they in the same single vlan with IP's on both side for a sudo layer-3 connection?

that will help. If it's layer-2 trunking I may have the anser for you. Also (if it's layer-2) when you notice the TCN does one end go to "inactive" and that's why you have to bounce the interface?

thanks,
Lee

 

[mjb22]

Thanks Lee,

In response to your question........

It is layer 2 trunking and no one end doesn't go inactive, it's always displayed as fast ethernet up, line protocol up at either end.

Thanks

 

[bierhunter]

You may want to double check some of your spantree settings on switches. A topology change can cause spantree algorithms to kick in and cause about 50 seconds of inactivity.

Make sure that all ports with a host (workstation, server) have portfast on. This will stop topology change notifications from happening there. Keep portfast OFF for uplink ports unless...you have a switch that has only one (and only one) path to it, you may turn on portfast for its uplink port. NOTE: This is not recommended, because you may add a redundant path later and forget that portfast is on. That can cause spantree loops...ick. But....if you know for a fact that you have no redundant connections to a switch, then turning portfast on for its uplink port will also help stop topology change notifications.