Latency on ICMP traffic on NSX Edge appliance's eth0

Products

• VMware NSX

Issue/Introduction

• NSX Manager alarms with "Network Latency High between Manager Node and Transport Node".
• The alarm triggers against NSX Edge appliance Virtual Machine.
• When doing a ping to the Edge's eth0, an intermittent increase in round-trip time is observed, such as a ping reply showing approximately 1946 ms:
  
  64 bytes from <NSX_Edge_eth0_IP>: icmp_seq=1234 ttl=61 time=1946 ms
  
  
• Packet captures on Edge's eth0 show that the latecy is introduced between the time when ICMP Echo Request is received on Edge's eth0 and a Reply is sent on the same interface, e.g.:
  
  11:12:10.559632 IP NSX_Manager  > NSX_Edge        ICMP echo request, id 2345, seq 1234, length 64 
11:12:12.505235 IP NSX_Edge     > IP NSX_Manager  ICMP echo reply, id 2345, seq 1234, length 64

A delay of nearly 2 seconds between receiving the ICMP request and sending the reply is observed, indicating processing latency within the Edge appliance.

Environment

VMware NSX 4.1.2.3

Resolution

If this issue is observed, impacted appliance can be redeployed via the following methods:

• Redeploy an NSX Edge VM Appliance, or
• Replacing an NSX Edge Transport Node in an NSX Edge Cluster.

If post redeploy this issue is not resolved please open a ticket with Global Support team and reference this KB article. 
For more information, see Creating and managing Broadcom support cases.