Troubleshooting

Symptom: Noticeable slowdown in some multicast traffic

If the switch is supporting more than 1022 active flows, this generates the message Unable to learn HW IP multicast groups, table FULL in the Event Log, because there is no room in the hardware MRT to add another multicast group. Software will route any multicast packets sent to multicast groups that are not in the hardware MRT, but it will be slower, and packets may be dropped if the data rate is greater than 3000 packets per second. See Flow capacity.


[NOTE: ]

NOTE: The PIM protocol uses oneMRT entry for every IP multicast S/G pair that it is routing. An entry is not used if the multicast flow is bridged and not routed. Entries in this table are automatically aged-out if they are unused for a period of time.


Heavy memory usage

Heavy use of PIM (many S/G flows over many VLANs), combined with other memory-intensive features, can oversubscribe memory resources and impact overall performance. If available memory is exceeded, the switch drops any new multicast flows and generates appropriate Event Log messages. Corrective actions can include:

  • Reducing the number of VLANs on the switches by moving some VLANs to another device.

  • Freeing up system resources by disabling another, non-PIM feature.

  • Moving some hosts to another device.

For more information, see Operating notes and Messages related to PIM operation.

IPv4 table operation

The IPv4 table, which contains the active IP multicast addresses the switch is currently supporting, has 128k entries. However, the IPv4 table also contains IP host entries for every IP source or destination that the switch has learned, as well as ACL flow entries. Entries in this table are generally aged-out if they are unused for 5 minutes or more.