Connectivity issue when HCX Network Extension (NE) are using Sink port
search cancel

Connectivity issue when HCX Network Extension (NE) are using Sink port

book

Article ID: 407556

calendar_today

Updated On:

Products

VMware HCX VMware vSphere ESXi

Issue/Introduction

Packet loss may occur when you've configured bridging ( HCX network extension) when the following conditions are met:

  • You are using DVS version 6.5 or earlier, OR DVS is above 6.6 but HCX was deployed before the DVS switch was upgraded, creating a sink port instead of a mac-learning port
  • From the "net-dvs -l" output where the NE appliance is running, you confirmed that the NE appliance is configured as SINK:
    com.vmware.etherswitch.port.extraEthFRP =   SINK
  • "MAC address changes & Forged transmits" policies under DVS port-group are in Reject state. For more information, access: HCX-NE:Considerations of "MAC address changes & Forged transmits" policies under DVS port-group
  • The sink bit (0x10000) is missing:
    • Get portset name and port number by running net-stats -l
    • Run the following command and check whether 0x10000 is set or not:
      vsish -e get /net/portsets/$switch_name/ports/$port/status | grep -A 2 "accepted:filter" | grep flags

            accepted:filter {
               flags:0x0000000d <<<<----- it should be: 0x0001000d
               unicastAddr:00:50:##:##:##:##:
               numMulticastAddresses:0
               multicastAddresses:
               LADRF:[0]: 0x0 
               [1]: 0x0 

      Note: Expected output is 0x0001000d; if you are seeing the flag 0x0000000d, you are matching this KB.
  • Using the command --trace on the "uplink" interface, you observe that the packet is dropped on EtherswitchFwdCheckPolicy, command is similar to:
    pktcap-uw --uplink vmnic# --srcip <source-ip> --dstip <destination-ip> --trace
      
      
            PATH:
              +- [11:46:44.230333] |                           VnicTx |   ######## |
              +- [11:46:44.230334] |                        PortInput |   ######## |
              +- [11:46:44.230334] |                          IOChain |            | VLAN_InputProcessor@com.vmware.vswitch#1.0.0
              +- [11:46:44.230335] |               EtherswitchDispath |   ######## |
              +- [11:46:44.230335] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230335] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230343] |                             Drop |    ########|
              +- [11:46:44.230343] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230343] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230344] |                             Drop |            |
              +- [11:46:44.230344] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230345] |                             Drop |            |
              +- [11:46:44.230345] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230346] |                             Drop |            |
              +- [11:46:44.230346] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230347] |                             Drop |            |
              +- [11:46:44.230347] |        EtherswitchFwdCheckPolicy |   ######## |
              +- [11:46:44.230348] |                             Drop |            |

Cause

The sink flag is removed from the ESXi host switchport, causing the packet to be dropped on the Uplink interface and never reach the NE switchport.

Resolution

This issue is resolved in VMware ESXi  8.0 U3, available at Broadcom downloads.
If you are having difficulty finding and downloading software, please review the Download Broadcom products and software KB.

Workaround :

Configure "MAC address changes & Forged transmits" to "Accept" state as per HCX-NE:Considerations of "MAC address changes & Forged transmits" policies under DVS port-group