Severe latency due to Unicast Probe & VDR Hold Queue Issue

Products

• VMware NSX

Issue/Introduction

Users experience severe latency when browsing.

1. Pick a couple of hosts that have Server and Client VMs deployed and execute the below command
   
   net-vdr -I -l
   
   
2. In the output look for the value with the arrow next to it. If it's 1000, use this host for running tests.
   
   Example output
   
   ==
   [root@localhost:~] net-vdr -I -l
   DR Instance Information :
   ---------------------------
   DR UUID: 2478a1e5-####-####-####-6ea3a0176f84
   DR Id: 0x00000002
   Number of Lifs: 3
   Number of Routes: 4
   Number of Hold Pkts: 1 <========
   State: Enabled
   Num unique nexthops: 1
   Number of ARP Entries: 18
   Max number of ARP Entries: 50000
   Generation Number: 0
   Edge Active: No
   Pmac: 00:00:00:00:00:00
   Dynamic resource pool tag: 65535
   Multicast Routing: Disabled
   ==
   
   
3. Execute the below command to find out LIF that has more Nascent ARPs. Note the UUID from the last column in the output.
   
   net-vdr -N -l <DR-UUID> | grep “00:00:00:00:00” > /tmp/vdr-neighbor-entries.txt
   
   ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
   created a vdr-neighbor-entries.txt file for all hosts with the 1000 Hold Packets (with the hostname in front)
   ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
   
   
4. Confirm the issue by identifying the hostnames with the 1000 Hold Packets
   
   less vdr-neighbor-entries.txt after running above command or
   
   net-vdr -N -l <DR-UUID>
   
   

Environment

VMware NSX-T Data Center 3.x

Cause

‘Number of Hold Packets’ counter never decrements.
Unicast ND probe not working.

Resolution

This issue is resolved in VMware NSX 3.2.0, available at Broadcom downloads.

If you are having difficulty finding and downloading software, please review the Download Broadcom products and software KB.