layout-3layout-6layout-1layout-top
NVIDIA Docs Hubchevron_right NVIDIA Networkingchevron_right Networking Softwarechevron_right Management Softwarechevron_right

NVIDIA NetQ

layout-toplayout-6layout-1

RoCE

Use the UI or CLI to monitor RDMA over Converged Ethernet (RoCE) for Spectrum switches and BlueField DPUs.

RoCE Commands

View RoCE Counters Networkwide in the UI

RoCE switches displays transmit (Tx) and receive (Rx) counters as well as counter pools for all switches running RoCE in your network. RoCE DPUs displays physical port, priority port, RoCE extended, RoCE, and peripheral component interconnect (PCI) information for all DPUs running RoCE in your network.

  1. From the header or Menu, select Spectrum-X, then RoCE.

  2. Select either RoCE switches or RoCE DPUs. The following displays a list of all switches running RoCE across a network.

View RoCE Counters for a Given Switch

You can view the following RoCE counters for a given switch:

  • Receive and transmit counters
  • General, CNP, and RoCE-specific counters
  • Counter pools
  • Port-specific counters

To view RoCE counters on a switch, search for the device’s hostname in the global search field or from the header select Add card > Device card. Select a switch that is running RoCE and open the large card on your workbench. Click the RoCE tab to view RoCE counters and their associated ports:

switch card displaying RoCE transmit nd receive data

Expand the card to the largest size, then select RoCE counters from the side menu. Use the controls above the table to view, filter, or export counter statistics by Rx, Tx, or Pool.

Disable RoCE Monitoring

To disable RoCE monitoring:

  1. (Optional) Verify that the NetQ agent is monitoring the RoCE service with netq config show agent commands. The Active column displays a yes status for the following services if RoCE monitoring is enabled:
cumulus@switch:~$ sudo netq config show agent commands
Service Key               Period  Active       Command                                                        Timeout
-----------------------  --------  --------  --------------------------------------------------------------  ---------
roce                           60  yes        Netq Predefined Command                                         None
roce-config                   300  yes        Netq Predefined Command                                         None
nvue-roce-config              300  yes        Netq Predefined Command                                         None
  1. Disable each of the three RoCE service keys individually with netq config add agent command service-key:
cumulus@switch~$ sudo netq config add agent command service-key roce enable False
Command Service roce is disabled
cumulus@switch:~$ sudo netq config add agent command service-key roce-config enable False
Command Service roce-config is disabled
cumulus@switch:~$ sudo netq config add agent command service-key nvue-roce-config enable False
Command Service nvue-roce-config is disabled
  1. Verify that NetQ is no longer monitoring the RoCE service with netq config show agent commands. The Active column displays a no status for the disabled services:
cumulus@switch:~$ sudo netq config show agent commands
Service Key               Period  Active       Command                                                        Timeout
-----------------------  --------  --------  --------------------------------------------------------------  ---------
roce                           60  no         Netq Predefined Command                                         None
roce-config                   300  no         Netq Predefined Command                                         None
nvue-roce-config              300  no         Netq Predefined Command                                         None

    Copyright © 2025  NVIDIA Corporation