4948E switch flushing MAC address table

[ffacilities]

Hi all,

I have an issue with a 4948E switch that's being used as a core switch flushing its MAC address table, seemingly at random. When this happens, we get ~5 second floods of unicast traffic, as you'd expect, which cripples the network.

The LAN is set up with 128 VLANs that all map into MST instance 1. Most switches are uplinked either directly into the core 4948E, or else into a 3750 stack that's connected via a 2Gb port-channel to the 4948E.

There doesn't seem to be any consistency to when the flushes happen. We can sometimes go a couple of minutes between them, then other times it'll happen twice in the space of 30 seconds. Immediately before the flush, the MAC address table is up around 2000 entries, which should be fine given the switch supports up to 55,000 entries.

Does anyone have any idea what might cause a switch to flush its MAC address table? I've already tried turning on spanning tree event debugging, but there's nothing that coincides with the flushes.

Thanks in advance!

Sean

[ffacilities]

I should also add, the switch is running IOS 12.2(54)SG1, but I couldn't see anything in the bug tracker that looked relevant.

[Danilo Dy · Accepted solution]

You may have a unicast flooding problem due to changes in your spanning-tree topology.

Try this http://www.ciscopress.com/articles/article.asp?p=336872

[ffacilities]

Thanks, I think you're right - I've just turned on 'debug span mst tc' and I'm seeing topology changes that coincide with the flushes. It looks like a lot are misconfigured trunk ports to end hosts without 'span portfast trunk' configured, so I'm going to work my way around those. Thanks for the tip!