[c-nsp] ARP table corrupted?

[Ruben Montes]

Hello,
 
I'm suffering a weird problem.
 
I'm trying to ping the ip address of two switches (c2950-12.1(22)EA1) that are conected in the same vlan that I am in and I get no response until I do a clear arp in the switches. Then I can ping them again, but if I wait some minutes I get the same problem and I have to clear the arp table. They are in the same vlan, so no L3 issues involved.
The diagram:
 
core1 -----switch1-----switch2
                     |
core2--------
 
Due to the nature of the connections, I think is a problem between the core and the switches. I have these outputs in switch1:
 
switch1#sh int trunk
Port        Mode         Encapsulation  Status        Native vlan
Fa0/1       on           802.1q         trunking      1
Fa0/2       on           802.1q         trunking      1

Port      Vlans allowed on trunk
Fa0/1       1,45,55,73,254,1002-1005
Fa0/2       1,45,55,73,254,1002-1005

Port        Vlans allowed and active in management domain
Fa0/1       1,45,55,73,254
Fa0/2       1,45,55,73,254

Port        Vlans in spanning tree forwarding state and not pruned
Fa0/1       1,45,254
Fa0/2       55,73

Fa0/1 connects with core1 and fa0/2 with core2. Vlan with problems is 254. This side is normal (spanning-tree is forwarding vlan 254 in Fa0/1, Fa0/2 is blocking) but in core1 side I have:
 
core1#sh int fa1/3 trunk
Port      Mode         Encapsulation  Status        Native vlan
Fa1/3     on           802.1q         trunking      1
 
Port      Vlans allowed on trunk
Fa1/3     1,45,55,73,254
 
Port      Vlans allowed and active in management domain
Fa1/3     1,45,55,73,254
 
Port      Vlans in spanning tree forwarding state and not pruned
Fa1/3     1
 
The strange thing here is last line: it seems to be pruning vlan 254 for that connection!!! Switch1 and 2 are vtp transparent but obviously have vlan254 configured and core1 shouldn't prune vlan 254 in this connection. I think this is the problem. It is not a configuration issue; I had a problem similar in the past that was solved doing a shut/no shut of the interface. Anyway, as you can see, vlan 45 is also being prunned and there are clients there: I cannot explain why they're not affected...!
 
What do you think?
 
Regards,
 
Ruben