Cisco Cisco IOS Software Release 15.2(3)E

Changing the load-balance distribution method or changing the number of ports in the EtherChannel can resolve this problem.

Use any of these workarounds to improve EtherChannel load balancing:

For random source-ip and dest-ip traffic, configure the load-balance method as src-dst-ip.

For incrementing source-ip traffic, configure the load-balance method as src-ip.

For incrementing dest-ip traffic, configure the load-balance method as dst-ip.

Configure the number of ports in the EtherChannel so that the number is equal to a power of 2 (that is, 2, 4, or 8)

For example, with load balance configured as dst-ip with 150 distinct incrementing destination IP addresses, and the number 
of ports in the EtherChannel set to either 2, 4, or 8, load distribution is optimal. (CSCeh81991)

When the rate of received DHCP requests

exceeds 2,000 packets per minute for a long time, the response time might be slow 

when you are using the console.

The workaround is to use rate limiting on DHCP traffic to prevent a denial of service attack from occurring. (CSCeb59166)

Some switch queues are disabled if the buffer size or threshold level is set too low with the mls qos queue-set output 
global configuration command. The ratio of buffer size to threshold level should be greater than 10 to avoid disabling the 
queue. 

The workaround is to choose compatible buffer sizes and threshold levels. (CSCea76893)

When auto-QoS is enabled on the switch, priority queuing is not enabled. Instead, the switch uses shaped round robin 
(SRR) as the queuing method. The auto-QoS feature is designed on each platform based on the feature set and hardware 
limitations, and the queuing method supported on each platform might be different. 

There is no workaround. (CSCee22591)

Cisco Discovery Protocol (CDP), VLAN Trunking Protocol (VTP), and Port Aggregation Protocol (PAgP) packets 
received from a SPAN source are not sent to the destination interfaces of a local SPAN session. 

The workaround is to use the monitor session session_number destination {interface interface-id encapsulation 
replicate} global configuration command for local SPAN. (CSCed24036)

IP traffic with IP options set is sometimes leaked on a trunk port. For example, a trunk port is a member of an IP multicast 
group in VLAN X but is not a member in VLAN Y. If VLAN Y is the output interface for the multicast route entry assigned 
to the multicast group and an interface in VLAN Y belongs to the same multicast group, the IP-option traffic received on 
an input VLAN interface other than the one in VLAN Y is sent on the trunk port in VLAN Y because the trunk port is 
forwarding in VLAN Y, even though the port has no group membership in VLAN Y. 

There is no workaround. (CSCdz42909).

For trunk ports or access ports configured with IEEE 802.1Q tagging, inconsistent statistics might appear in the show 
interfaces counters privileged EXEC command output. Valid IEEE 802.1Q frames of 64 to 66 bytes are correctly 
forwarded even though the port LED blinks amber, and the frames are not counted on the interface statistics.

There is no workaround. (CSCec35100).