[c-nsp] UDLD enabling port prematurely?

[Victor Sudakov]

Colleagues,

I have several pairs of Catalyst 3560E switches connected via third
party MUXes (Avara ENT100). When the actual physical medium goes down,
the MUXes do not shutdown their Ethernet interfaces (i.e. they have no
"link poisoning"). So I need some sort of point-to-point L2 link fault
management between the switches.

Is UDLD suitable for this purpose? I have experimented a bit with
"udld port aggressive" and have found out the following strange
thing.

When the physical link goes down, UDLD detects this condition and
shuts the switch interface down. However, after several minutes, the
interface goes up again with "%PM-4-ERR_RECOVER: Attempting to recover
from udld err-disable state on Gi0/17". The interface is up even
though "Current bidirectional state: Unknown", and seems to be in the
STP forwarding state.

This does not look good. If a broadcast frame arrives to this port
before a BPDU does, there will be a storm and a lot of MAC flapping.
Is there a way to keep the port from forwarding traffic until the UDLD
state is Bidirectional ?

Thanks a lot for any input.

-- 
Victor Sudakov,  VAS4-RIPE, VAS47-RIPN
sip:sudakov at sibptus.tomsk.ru