Administrator Guide

Propagation of DCB Information
When an auto-upstream or auto-downstream port receives a DCB conguration from a peer, the port acts as a DCBx client and
checks if a DCBx conguration source exists on the switch.
If a conguration source is found, the received conguration is checked against the currently congured values that are internally
propagated by the conguration source. If the local conguration is compatible with the received conguration, the port is
enabled for DCBx operation and synchronization.
If the conguration received from the peer is not compatible with the internally propagated conguration used by the
conguration source, the port is disabled as a client for DCBx operation and synchronization and a syslog error message is
generated. The port keeps the peer link up and continues to exchange DCBx packets. If a compatible conguration is later
received from the peer, the port is enabled for DCBx operation.
NOTE: DCB congurations internally propagated from a conguration source do not overwrite the conguration on a
DCBx port in a manual role. When a conguration source is elected, all auto-upstream ports other than the conguration
source are marked as
willing disabled
. The internally propagated DCB conguration is refreshed on all auto-conguration
ports and each port may begin conguration negotiation with a DCBx peer again.
Auto-Detection and Manual Conguration of the DCBx Version
When operating in Auto-Detection mode (the DCBx version auto command), a DCBx port automatically detects the DCBx
version on a peer port. Legacy CIN and CEE versions are supported in addition to the standard IEEE version 2.5 DCBx.
A DCBx port detects a peer version after receiving a valid frame for that version. The local DCBx port recongures to operate with
the peer version and maintains the peer version on the link until one of the following conditions occurs:
The switch reboots.
The link is reset (goes down and up).
User-congured CLI commands require the version negotiation to restart.
The peer times out.
Multiple peers are detected on the link.
If you congure a DCBx port to operate with a specic version (the DCBx version {cee | cin | ieee-v2.5} command in
the Conguring DCBx), DCBx operations are performed according to the congured version, including fast and slow transmit timers
and message formats. If a DCBx frame with a dierent version is received, a syslog message is generated and the peer version is
recorded in the peer status table. If the frame cannot be processed, it is discarded and the discard counter is incremented.
NOTE: Because DCBx TLV processing is best eort, it is possible that CIN frames may be processed when DCBx is
congured to operate in CEE mode and vice versa. In this case, the unrecognized TLVs cause the unrecognized TLV
counter to increment, but the frame is processed and is not discarded.
Legacy DCBx (CIN and CEE) supports the DCBx control state machine that is dened to maintain the sequence number and
acknowledge the number sent in the DCBx control TLVs.
DCBx Example
The following gure shows how DCBX is used on an MXL Switch installed in a PowerEdge M1000e chassis in which servers are also
installed.
The external 40GbE ports on the base module (ports 33 and 37) of two switches are used for uplinks congured as DCBx auto-
upstream ports. The MXL switch is connected to third-party, top-of-rack (ToR) switches through 40GbE uplinks. The ToR switches
are part of a Fibre Channel storage network.
The internal ports (ports 1-32) connected to the 10GbE backplane are congured as auto-downstream ports.
On the MXL switch, PFC and ETS use DCBx to exchange link-level conguration with DCBx peer devices.
228
Data Center Bridging (DCB)