NSX Edge HA not working as expected

Products

• VMware NSX

Issue/Introduction

• NSX Edge High Availability fails.
• It may be seen that after all BFD sessions going down the Standby Edge became Active, but when the BFD sessions are restored both Edges become Active.
• Below logs may be seen:

  • At Edge node log file /var/log/syslog : 
    Edge## NSX #### FABRIC [nsx@6876 comp="nsx-edge" subcomp="datapathd" s2comp="ha-transport-#####" tname="dp-bfd-###" level="WARN"] MC-#####: rx bind on ###########:50263 fail: Address already in use ###

Environment

VMware NSX

VMware NSX-T Data Center

Cause

The UDP port 50263 is used by another process and this prevent NSX Edge to exchange HA control messages via UDP port 50263.

NSX Edge use UDP port 50263 to exchange control messages with each other. This port is not reserved and can be occupied by other UDP client if Linux kernel happens to allocate this port to another process. This will result in HA control messages not be able to be exchanged between NSX Edges therefore the HA will not work as expected.

Resolution

This issue has been resolved in NSX 4.2.3.

Workaround:

Reserve UDP port 50263 so Linux kernel will not allocate it to other processes.

sysctl -w net.ipv4.ip_local_reserved_ports=50263 >> /etc/sysctl.conf