3548 switch with lots of underruns

[2tramsay]

I have a branch office with five 3500 series switches - all with 48 ports. Two 3550's and three 3548's. They are interconnected with Gigastack modules and SX GBIC's. I have one 3548 switch that seems to be reporting a lot of underruns on all the FE ports. Everything checks out from what I can tell. No buffer problems, etc.. I've reset the switch, but the underruns are still occuring. Attached is a show interface text document and a diagram of the environment. Any ideas would be greatly appreciated!! (Tor3548-1B is the switch in question)

[a-vazquez]

The HSSI network module has two serial controllers. One controller is used for transmit packets, and the other for receive packets.The HSSI bus arbitration scheme gave preference to the transmit controller. This could lead to the receive controller being unable to access the bus in order to transfer packets into memory, leading to an overrun (which results in the packet being dropped).

The bus arbitration scheme gives equal priority to the receive controller. This will reduce the number of receive overflows. However, since the transmit controller has lower priority that before, it may increase the number of transmit underruns.

[mathieunantel]

I have the exact same problem with one of my customers. The exact same setup, with gigastack modules, and a pair of 3550 switchs at the core! I have between 5000 and 100000 underruns on the ports of all of the 3548s. I suspect this problem might be responsible for some specific app disconnections. I'm currently trying to put an affected user on one of the 3550 to verify whether this has any impact on performance problems. Any answer to the OP's problem might just as well solve mine...

Thanks!

Mat