We have recently had a problem with a number of 2950 switches (various models/IOS) where on occasion our monitoring stations, which run various software and sit in different locations/subnets, will lose polling contact with the switch management address.
On every such occasion I can see ICMP/SNMP packets coming into the switch via an ACL log, but whenever I try to send either a ping or an SNMP trap from the switch the packet doesn't appear to leave the switch. This isn't a routing issue as I can ping other hosts in the subnet that house the management stations from the switch and vice versa. Its as if the switch just gives up sending packets to the specific destination?
The only commonality in the config is that the IP of the management station is within the 'logging host <ip>' statement.
The only way I have found to clear the problem is to restart the switch. Has any one seen this before? We don't see it on non 29** switches.