Switches "locking" up

[timpotter]

Okay, we've had this happen twice now, so it was not a fluke. Here's the situation:

We have 3 Cat 3524XL switches all running the same IOS 12.0(5)WC7. We also have 2 Cat 3550-48 switches running 12.1(11) and 12.1(13) IOS.

The are all hooked together in a cascaded redundant loop fashion with gbic cables and modules. One of the switches has a fiber uplink to another closet.

Well, yesterday at about 4:50PM, the switches all appeared to lock up - or maybe go into a loop? One of the GBIC ports was still blocked, so I don't think it was a loop. We lost all network connectivity for that closet and floor. Other closets were unaffected.

When I unplugged a GBIC cable and plugged it back in again, the switches came back alive and all was good.

I suppose putting both 3550's to the same IOS might be the fix, but I'm not convinced. If that is the problem, it sure has a strange way of showing itself and fixing itself. And we've been running this way for months.

Any other options as to what might be wrong? I already verified that all ports had portfast disabled.

Thanks,

Tim

[konigl]

Sounds like you've got an Ethernet loop here, or maybe a GigaStack configuration problem.

Are you using GigaStack GBICs to create the cascaded redundant loop connecting the three 3524XL and two 3550-48 switches?

Is there only one Gigabit fiber uplink from that closet to another closet? Or do you have redundant fiber connections as well?

Do you have multiple VLANs assigned to ports across the GigaStack of switches in that closet? Which ports are configured as VLAN trunk ports?

If you are running GigaStack GBICs, here's a useful link:

Catalyst Switch GigaStack Configuration and Implications

http://www.cisco.com/en/US/products/hw/switches/ps607/products_tech_note09186a00800a2cac.shtml

[jdmckeever]

I would debug spanning tree events on the 3550s. It will log topology change notices sent and received by the switch and on which interface - a useful debugging tool.