Forwarding(?) problem on 2848

[Terry Kirk]

We have two 2848s that connect our servers. The two 2848s are then linked to a pair of 5308s that are the core switches.

I have seen the problem once on each of the 2848s.

A server connected to the 2848 with the problem will drop off of the network (IP).

It is able to ping the 2848 it is connected to and the 2848 can ping the server. Switches further down the chain, like the core switches can ping the server, but the server can't ping them. The server does have correct arp entries for the IP addresses it is trying to ping.
This problem seems to affect only a few of the servers attached to the switch.

MAC table entries seem correct. PCM+ 2 reports different paths depending on which way I put the start and end nodes in, but I'm not sure how it scans exactly.

Rebooting the server has no affect.

Rebooting the switch seems to fix the problem

The switches are all running the latest firmware.

Any ideas on how to chase this problem would be appreciated.

Terry

[Les Ligetfalvy]

You talk of everything in pairs but you fail to mention how they are connected. Are you trunking the servers to separate 2848 switches and are you routing to separate 5308 switches?
Are the core switches meshed?
Do you have a drawing?
What code level are you running?