Rapidly pinging a vmk interface causes ping loss on ESXi

Products

• VMware vSphere ESXi

Issue/Introduction

• Rapidly pinging an ESXi host results in ping loss (like the ping command from a Cisco switch)
• The ping loss occurs after 200 pings per second
• Pinging at a slow interval does NOT experience ping loss (like the ping command from a Windows desktop)
• The issue only occurs on the vmk interfaces of the ESXi host and not on virtual machines

Environment

VMware vSphere ESXi 

Resolution

This is expected behavior. In version 6.0 of ESXi, a change was made to the TCP/IP stack to limit the number of ICMP responses per second. The response is limited to 200 pings per second. This value is hard coded and cannot be changed at this time.

The change was made for security purposes. It helps protect from denial of service attacks using ICMP.

This change does not impact the performance of ESXi or any traffic other than ICMP.

The ping loss can be reproduced by using the '-f' (flood) switch for the Linux ping command:

# ping -c 205 -f 10.21.22.160
PING 10.21.22.160 (10.21.22.160) 56(84) bytes of data.
......
--- 10.21.22.160 ping statistics ---
205 packets transmitted, 199 received, 2% packet loss, time 153ms
rtt min/avg/max/mdev = 0.335/0.405/1.059/0.089 ms, ipg/ewma 0.754/0.440 ms

NOTE: You must be root to use the '-f' switch and should use the '-c' (count) switch to limit the number of pings.

This also can be reproduced with vmkping command with the option -i for an interval below 0.005