Intermittent ICMP latency while pinging a DLR LIF or tracerouting through a DLR LIF

Products

• VMware NSX

Issue/Introduction

Symptoms:

• When pinging a DLR LIF or running a traceroute to view the datapath traversing a DLR LIF, you may see responses with higher latency. This latency can be in excess of 1000ms.
• Despite these periodic high latency ICMP responses, VM traffic traversing the same datapath does not experience this latency.

Environment

VMware NSX for vSphere 6.2.x
VMware NSX for vSphere 6.3.x
VMware NSX for vSphere 6.4.x

Resolution

This is expected behaviour.

ICMP frames sent from the DLR module in ESXi is queued in a different manner than actual traffic being routed. The queue used for ICMP will often be handled by a different processor thread and will sometimes need to awaken upon request. This can result in periodically slow ICMP responses.