Table of Contents
Advertisement
Command Modes
Global Configuration mode
User Guidelines
Unknown multicast and IPv4/IPv6 data packets destined to hosts in the
connected networks on the router for which the MAC address is not resolved
are trapped to CPU to trigger the ARP/neighbor discovery resolution of those
hosts.
When the ARP or neighbor table is filled, the switch cannot accommodate
new entries. In this case, there is no value in receiving the unresolved
IPv4/IPv6 packets. Likewise, in cases of a L2 network re-convergence, a large
number of neighbors may not be discovered but may be transmitting traffic.
In the case of multicast data, certain multicast topologies using multi-access
VLANs may result in packets being forwarded to the CPU with no associated
PIM or MFDB state.
Receiving large numbers unresolved packets spikes the CPU usage to high
levels at no benefit. For Ipv6, it also results in delayed processing of the NUD
packets (NS/NA) for the existing neighbor entries leading to NUD anomalies
and deletions of existing neighbor entries.
To avoid such an unnecessary CPU load leading to NUD anomalies when the
ARP or IPV6 neighbor table is close to full (crossing 95% of table size) or
other failures, the switch automatically reduces the rate limit to an empirical
value of 50 pps irrespective of the configured rate limit. When the table size
falls below 95% of the table size, it is restored to the configured rate limit
value.
Use this command to limit the CPU load in situations where large numbers of
unknown multicast or IPv4/IPv6 packets with an unknown multicast or
unicast IPv4/IPv6 destination are being handled in software. The symptom
can be diagnosed by high CPU usage of the ipMapForwardingTask.
Example
An example output is showing higher than normal CPU usage due to packets
copied to the software forwarding task below:
console#show process cpu
Memory Utilization Report
1028
Security Commands
Advertisement
Table of Contents
