vSAN File Services fail with reconnecting loop

Products

VMware vSAN 8.x VMware vSphere ESX 8.x

Issue/Introduction

After upgrading to VMware ESXi 8.0 Update 3, vSAN File Services may fail to operate correctly. The primary symptom is the vSAN File Service Virtual Machines (FSVMs) entering a continuous cycle of reconnecting or retrying attempts.

Symptoms:

Skyline Health score fluctuates significantly, typically between 97% and 99/100.
File Service health status reports "Service is (re)starting."
100% packet loss occurs when attempting to ping the gateway from the FSVM container.
DNS forward/reverse lookup and Domain Controller reachability errors are reported even when external networking is functional.

Environment

VMware vSAN ESA 8.x
ESXi Version: 8.0 Update 1 and higher (specifically Build 25595708)
Configuration: vSAN File Services with Express Storage Architecture (ESA) or OSA.
Networking: Port groups with Promiscuous Mode enabled.

Cause

This issue is caused by a conflict between the MAC management profile update introduced in ESXi 8.0u1 and specific virtual switch security settings. If Promiscuous Mode is enabled on the Distributed Virtual Switch (VDS) or Port Group, it may cause MAC Learning to be disabled at the host level for FSVM ports after a reboot or upgrade.

The vSphere Client UI may incorrectly display MAC Learning as "Enabled," but the host runtime configuration remains "False," preventing the FSVM from communicating properly.

Resolution

Review this command before running it.

Log in to the vSphere Client.
Navigate to the Port Group used for vSAN File Services.
Ensure Promiscuous Mode is set to Reject. If your environment requires it to be allowed, proceed to the CLI steps below to manually override the port state.
Log in to the affected ESXi host via SSH.
Run the following command to identify the dvPort ID for the file service node:
```
esxcfg-vswitch -l | grep -i "vsan file service node"
```
Verify the runtime status of MAC Learning for that port (replace #### with your port ID and DVS alias):
```
netdbg vswitch mac-learning port get --dvport #### --dvs-alias ####
```

If MAC Learning returns False, manually enable it and Unicast Flooding:

netdbg vswitch mac-learning port set --enable --flooding --dvport #### --dvs-alias ####

Return to Skyline Health and click Retest. The score should restore to 100%.

Additional Information

1. Fixed in ESXi 8.0 Update 3 and higher. See Download Broadcom products and software for steps to download the latest release.
2. To receive updates on this issue, please Subscribe to this Article. (https://knowledge.broadcom.com/external/article/275360)
3.Internal Tracking ID: PR 3308461 / Case 37028552