Cisco Support Community

Catalyst 6500 switch having intermittent connectivity issues until the ARP cache is cleared

Core issue

Address Resolution Protocol (ARP) and adjacency tables are periodically refreshed independently. This may cause tables to be out of synch until this situation ages out.

Cisco Express Forwarding (CEF) is a Layer 3 (L3) IP switching technology that provides superior performance compared to other switching technologies, especially in networks with dynamic traffic patterns.

CEF maintains data structures called Forwarding Information Base (FIB) and adjacency tables. The FIB table mirrors the information in the routing table and is used to make forwarding decisions. The adjacency table contains the precomputed link-layer header for next hop devices. Based on the next hop interface, entries in the FIB table are mapped to entries in the adjacency table. A device is not able to CEF switch packets if the adjacency table is not populated with the required information.

If CEF drops packets at regular intervals, interspaced by periods of normal operation, it is probably due to the adjacency table being cleared periodically. This is caused by the aging of the Address Resolution Protocol (ARP) entry. Packets are not CEF switched for the duration in which the adjacency table is repopulated with the required next hop information. While ARP entries are refreshed by default every four hours, configuring a very small value of ARP timeout is disruptive to CEF operation.

This is also identified in Bug:CSCeb53542


Configure the ARP timeout to be 60 seconds or a multiple of 60 seconds. For  example, when you enter the arp timeout 270 interface configuration command, the symptom occurs. However, when you enter the arp timeout 300 interface configuration command, the symptom does not occur.

For more information, refer to