Preview Tool

Cisco Bug: CSCvu15037 - Cisco Nexus 9000 Cloud Scale devices drop 25G broadcast traffic as output discard with queue-limit

Last Modified

Jun 24, 2020

Products (66)

  • Cisco Nexus 9000 Series Switches
  • Cisco Nexus 3548 Switch
  • Cisco Nexus 9516 Switch
  • Cisco Nexus 93600CD-GX Switch
  • Cisco Nexus 3548-X Switch
  • Cisco Nexus 92348GC-X Switch
  • Cisco Nexus 3636C-R Switch
  • Cisco Nexus 9396TX Switch
  • Cisco Nexus 31108TC-V Switch
  • Cisco Nexus 93108TC-FX Switch
View all products in Bug Search Tool Login Required

Known Affected Releases

7.0(3)I7(6) 7.0(3)I7(8) 9.3(4)

Description (partial)

A Cisco Nexus 9000 switch with the Cloud Scale ASIC will drop multi-destination/BUM (Broadcast/Unknown Unicast/Multicast) traffic as an output discard if the interface's egress queuing policy has a static queue-limit defined for the traffic's queue. Since the affected traffic is multi-destination in nature, output discards will increment on all interfaces that are forwarding the ingress VLAN in question.

As this issue affects all multi-destination traffic, this issue can manifest itself with multiple different symptoms. Some common ones include:

+++ 25G-connected hosts are not able to obtain IP addresses via DHCP
+++ 25G-connected hosts are not able to establish network connectivity through ARP resolution
+++ 25G-connected hosts are not able to receive multicast traffic via IGMP

1. Nexus 9000 device with a Cloud Scale ASIC.
2. Multidestination/BUM (Broadcast/Unknown Unicast/Multicast) traffic ingresses the Nexus 9000 through a 25G interface.
3. A non-default system-wide (or interface level) egress queueing policy is applied to the device, such as with the following configuration:

N9K# show running-config ipqos
system qos
  service-policy type queuing output CUSTOM-Q-POLICY

4. The egress queuing policy in question has a static queue-limit defined for the queue where the affected multidestination traffic would be classified, such as with the following configuration.

N9K# show policy-map type queuing CUSTOM-Q-POLICY
  policy-map type queuing CUSTOM-Q-POLICY
    class type queuing c-out-q3
      priority level 1
      queue-limit 100 kbytes              <<<
    class type queuing c-out-q2
      queue-limit 100 kbytes              <<<
    class type queuing c-out-q1
      queue-limit 100 kbytes              <<<
    class type queuing c-out-q-default
      queue-limit 200 kbytes              <<<

5. Incrementing output discards are observed on all interfaces where the VLAN of the ingress traffic is being forwarded. For example, consider multidestination traffic that ingresses Ethernet1/1 in VLAN 25 placed in the default queue. Ethernet1/2 and Ethernet1/3 are also forwarding VLAN 25. Incrementing output discards will be observed on Ethernet1/2 and Ethernet1/3, even if both Ethernet1/2 and Ethernet1/3 are not oversubscribed with traffic.

N9K# show running-config interface Ethernet1/1
  switchport mode trunk
  switchport trunk allowed vlan 25

N9K# show spanning-tree vlan 25
Interface        Role Sts Cost      Prio.Nbr Type
---------------- ---- --- --------- -------- --------------------------------
Eth1/1           Desg FWD 800       128.57   P2p 
Eth1/2           Desg FWD 500       128.58   P2p 
Eth1/3           Desg FWD 500       128.59   P2p 

N9K# show interface counters errors non-zero 
Port          Align-Err    FCS-Err   Xmit-Err    Rcv-Err  UnderSize OutDiscards
Eth1/2                0          0          0          0          0       74283
Eth1/3                0          0          0          0          0       74283
Bug details contain sensitive information and therefore require a account to be viewed.

Bug Details Include

  • Full Description (including symptoms, conditions and workarounds)
  • Status
  • Severity
  • Known Fixed Releases
  • Related Community Discussions
  • Number of Related Support Cases
Bug information is viewable for customers and partners who have a service contract. Registered users can view up to 200 bugs per month without a service contract.