docs: Clarify DCBX requirement vs DCBX willing mode in TOR QoS reference

[ebmarquez]

Summary

The TOR QoS Policy reference doc creates customer confusion about DCBX configuration. Customers are making a logical leap from "DCBX must be enabled to advertise the required TLVs" to "DCBX willing mode should be true." This is incorrect. Azure Local does not support DCBX willing mode — only a static configuration is supported.

Affected doc: Reference-TOR-QOS-Policy-Configuration.md

Referenced from: Physical network requirements

Customer Confusion

The current language describes enabling DCBX as a required aspect of switch configuration. This is true in the sense that DCBX must be enabled to advertise all the TLV values required for Azure Local (carried in LLDP frames). However, customers are inferring that DCBX willing mode should also be enabled.

What's Correct

• DCBX enabled: required to advertise the necessary TLVs over LLDP.
• DCBX willing mode: not supported in Azure Local. Configuration must be static.

Why DCBX Willing Mode Is Not Supported

DCBX willing mode allows the fabric to dynamically change host-level settings. The fabric does not understand Azure Local's RDMA / storage requirements, and dynamic changes from the fabric can disrupt RDMA traffic and impact the storage layer. If DCBX willing mode were enabled, a rogue configuration could change neighbor settings and break system configurations.

This is not a new requirement — it has been in place since Storage Spaces Direct originally launched.

Requested Change

Update the TOR QoS reference doc to:

1. Explicitly state that DCBX willing mode must be disabled / false (static configuration only).
2. Distinguish between "DCBX enabled (for TLV advertisement)" and "DCBX willing mode (not supported)".
3. Briefly explain why willing mode is not supported (RDMA / storage traffic protection), so customers understand the rationale and don't try to enable it.

Scope

Small documentation fix. No behavior change.