PSOD on Hosts with DLB and DFW configured on NSX 4.2.2.1 version

Products

• VMware NSX

Issue/Introduction

PSOD on ESXi hosts can occur with below functions in the backtrace, while Distributed Load balancer is configured on the hosts:

• pf_validate_session_v2
• pf_test_state_tcp 
• pf_test
• PFFilterPacket 
• VSIPFWProcessPackets
• VSIPDVFProcessPacketsInt

Environments with high traffic flows using DFW/ DLB and high change churn from DLB/DFW are prone to this issue.

Environment

VMware NSX

Cause

1. A DLB session is created based on the DLB and DFW configuration.
2. When DLB or DFW is reconfigured, the DLB session will be re-evaluated.
3. After revalidation, if the DLB session is not valid anymore, it will be freed in the purge thread.
4. Concurrently this session can be handled by packet handling thread, which results in host PSOD doe to this race condition.

Resolution

This will be fixed in the upcoming versions of NSX.

In case you encounter this issue, please reach out to Broadcom Support team.