Enabling Microburst Monitoring on Cisco Nexus Switches

Enabling Microburst Monitoring on Cisco Nexus Switches

Summary

This article explains how to enable and verify microburst detection on Cisco Nexus 9000 series switches.
Microbursts are short spikes of traffic that can momentarily exceed interface buffer capacity, leading to output discards even when average utilization is low.
By enabling burst monitoring, administrators can gain visibility into these transient congestion events.

Applies to:

  1. Cisco Nexus 9000 Series (NX-OS 9.x and later)
  2. Switches operating in 8-queue (8q) mode

Configuration Steps

  1. Create a Policy Map for Burst Detection
    1. Create a queuing policy that includes burst detection under the output queue

    2. The given below example uses "default-8q-out-policy" logic. You can adjust queue priorities or bandwidth based on your existing QoS setup

      policy-map type queuing MICRO-BURST-MONITOR

        class type queuing c-out-8q-q7

          priority level 1

        class type queuing c-out-8q-q6

          bandwidth remaining percent 0

        class type queuing c-out-8q-q5

          bandwidth remaining percent 0

        class type queuing c-out-8q-q4

          bandwidth remaining percent 0

        class type queuing c-out-8q-q3

          bandwidth remaining percent 0

        class type queuing c-out-8q-q2

          bandwidth remaining percent 0

        class type queuing c-out-8q-q1

          bandwidth remaining percent 0

        class type queuing c-out-8q-q-default

          bandwidth remaining percent 100

          burst-detect rise-threshold 90 percent fall-threshold 18 percent


  2. Apply the policy to Target Interfaces

    1. Apply the created queuing policy to all interfaces where you want to monitor for microbursts

      1. interface Ethernet1/1

          service-policy type queuing output MICRO-BURST-MONITOR

    2. Repeat the above step for each relevant interface

  3. Verify Configuration

    1. Use the following command to confirm the policy is applied:

      1. show run interface ethernet1/1

    2. You should see:

      1. service-policy type queuing output MICRO-BURST-MONITOR

  4. Monitor for Burst Events

    1. Once configured, monitor detected burst events using:

      1. show queuing burst-detect

    2. If a burst is detected, it will display the affected queue, rise/fall timestamps, and thresholds exceeded

  5. Expected Behavior

    1. During normal operation:

      1. While running show queuing burst detect it returns no active bursts

    2. During a Burst

      1. The command will show entries indicating which queues experienced congestion


Common Warnings

  1. "Egress policy with one or more class without priority/bandwidth"
  2. The above warning occurs when queues don't have bandwidth/priority explicitly defined
  3. Ensure bandwidth/priority statements exist for all queues, as in the sample above

Verification Commands

Command
Description
  show queuing burst detect
 Displays detected microbursts
  show interface counters errors
 Checks for output discards
  show policy-map interface ethernnet1/x
 Verifies applied policy