C2950 Switches *locking* up...

[PBroome51]

I'm currently running 15 Catalyst 2950s (which are cascaded in groups of 3 from a 12 (GB) port 3550). At random intervals in the day, one or more of the 2950s (different one each time) is just *locking up* (lights go solid) - the problem can only be rectified by restarting the switch.

This is starting to get somewhat annoying... Does anybody have any idea of anything that could be causing this. (Configuration-wise, I suspect...)

(The switches are all set to basic configuration, the only things set on them are individual IP addresses, host names, default gateway and DHCP server address). *Any* suggestions gratefully received...

[deilert]

My guess would be some kind of a spannning tree loop or a broadcast storm

[PBroome51]

Thanks, I'll look into that. A quick question as a followup, that may sound basic, but I'm kind of learning as I go here...

If enable PortFast on all the ports on the switches, will this stop the possible STP loops?

[deilert]

No , enabling port fast should be done only on ports that connect to server , PC , etc , never on ports connected to other switches or routers , enabling this forces the box to go directly into forwarding state skipping the STP algorithm .

[PBroome51]

Thanks again for that (you see I'm learning lots today...) One more configuration question. On the 3550 what VLAN mode should I be using, for basic usage? (Static Access/Trunk Desirable/Trunk Auto/Trunk/Trunk Non Negotiate...)

We just have a single VLAN set up here - no clever splitting of switches, etc... All the switches connected to the 3550 are part of the same subnet...

[rajesh444]

Trunk setting of "desirable" can be useful for troubleshooting if one side of the link drops versus "on" mode where the trunking is setup unconditionally but both can be used.

Also make sure that the management VLAN for your switches is seperate from user VLAN to isolate broadcast storms and ensure STP BPDU traffic is not affected.

Hope this helps,

RAJ

[PBroome51]

Many thanks for all the suggestions. It turns out the problem is being caused by a broadcast storm, emanating from a new Dell laptop with a Broadcom Gigabit NIC which a user has been testing on the network.

I should've guessed...