Reset Control

NVIDIA BlueField BMC Software v24.04
Note

Rebooting the DPU immediately after rebooting the DPU BMC is restricted. The user should wait until the IPMI service becomes operational before performing a DPU reboot, with a suggested delay of 30 seconds.

Issue the following command from the BMC to get the power status of the DPU:

Copy
Copied!
            

sudo curl -k -u root:'<password>' -H 'Content-Type: application/json' -X GET https://<bmc_ip>/redfish/v1/Systems/Bluefield/

Example output:

Copy
Copied!
            

{ … "PowerRestorePolicy": "AlwaysOn", "PowerState": "On", … }

Hard Reset of BlueField Arm Cores and NIC Subsystem

Info

Hard reset of the BlueField DPU is allowed only when the host asserts:

  • PERST signal on BlueField-2

  • All_STANDBY signal on BlueField-3

Copy
Copied!
            

curl -k -u root:'<password>' -H "Content-Type: application/json" -X POST https://<bmc_ip>/redfish/v1/Systems/Bluefield/Actions/ComputerSystem.Reset -d '{"ResetType" : "PowerCycle"}'

Example output:

Copy
Copied!
            

{ "@Message.ExtendedInfo": [ { "@odata.type": "#Message.v1_1_1.Message", "Message": "The request completed successfully.", "MessageArgs": [], "MessageId": "Base.1.15.0.Success", "MessageSeverity": "OK", "Resolution": "None" } ] }


Hard Reset of BlueField Arm Cores

Copy
Copied!
            

curl -k -u root:'<password>' -H "Content-Type: application/json" -X POST https://<bmc_ip>/redfish/v1/Systems/Bluefield/Actions/ComputerSystem.Reset -d '{"ResetType" : "ForceRestart"}'

Example output:

Copy
Copied!
            

{ "@Message.ExtendedInfo": [ { "@odata.type": "#Message.v1_1_1.Message", "Message": "The request completed successfully.", "MessageArgs": [], "MessageId": "Base.1.15.0.Success", "MessageSeverity": "OK", "Resolution": "None" } ] }


Soft Shutdown of BlueField Arm OS

Note

This command is relevant only for BlueField-3 DPUs.

Copy
Copied!
            

curl -k -u root:'<password>' -H "Content-Type: application/json" -X POST https://<bmc_ip>/redfish/v1/Systems/Bluefield/Actions/ComputerSystem.Reset -d '{"ResetType": "GracefulShutdown"}'

Example output:

Copy
Copied!
            

{ "@Message.ExtendedInfo": [ { "@odata.type": "#Message.v1_1_1.Message", "Message": "The request completed successfully.", "MessageArgs": [], "MessageId": "Base.1.15.0.Success", "MessageSeverity": "OK", "Resolution": "None" } ] }

Monitoring DPU OS Shutdown with Redfish

When the DPU OS shuts down successfully, PowerState changes to Paused and StatusState changes to StandbyOffline.

Copy
Copied!
            

curl -k -u root:'<password>' -H "Content-Type: application/json" -X GET https://<bmc_ip>/redfish/v1/Systems/Bluefield

Example output:

Copy
Copied!
            

… "PowerState": "Paused", … "Status": { "Health": "OK", "HealthRollup": "OK", "State": "StandbyOffline" }, …

BMC supports reset control of NVIDIA® BlueField® through the GPIOs connected to the BMC.

Issue the following command from the BMC to get the power status of the DPU:

Copy
Copied!
            

ipmitool chassis power status

To perform a reset of the DPU, use the following commands:

Description

Command

Hard reset of BlueField DPU (Arm cores and NIC)

Copy
Copied!
            

ipmitool chassis power cycle

Hard reset of BlueField Arm cores

Copy
Copied!
            

ipmitool chassis power reset

Soft Shutdown of BlueField Arm OS

Note

This command is relevant only for BlueField-3 DPUs.

Copy
Copied!
            

ipmitool power soft

Note

Hard reset of the BlueField DPU is allowed only when the host asserts:

  • PERST signal on BlueField-2

  • All_STANDBY signal on BlueField-3

Note

Soft shutdown of BlueField Arm OS is allowed only when the Arm OS is running. To retrieve the Arm OS state, refer to the 0xA3 command under "IPMItool NIC Subsystem Management".

Note

Between each reset control, there should be a wait until the system finishes the operation.

  • 20-second wait in BlueField-2

  • 5-second wait in BlueField-3

OEM command 0xA1 is defined for additional non-standard reset controls of BlueField from BMC under the OEM NetFn group 0x30.

NVIDIA OEM command to reset the BlueField:

Request

Response

Reset Option

  • 0x32 – NetFun

  • 0xA1 – command

  • 0x00 – Req_data1 (reset option)

Completion code:

  • 0x00 – success

  • <ipmi-error-code> – failure

  • 0x02 – soft reset of BlueField Arm cores

    Info

    This reset command is only available when the DPU OS is up.

  • 0x03 – reset on-board 3-port switch

Monitoring DPU OS Shutdown Using IPMI

After a successful shutdown, the DPU Arm enters a low-power standby state.

Info

The DPU Arm cannot be fully powered off, and Standby is its final state

To get the DPU's OS state, refer to the 0xA3 command under "IPMItool NIC Subsystem Management".

To get the DPU Arm to boot back to the DPU OS, users can either power cycle the DPU or perform a hard reset of the DPU Arm.

Info

The output of IPMItool chassis power status will show "Chassis power is on".


© Copyright 2024, NVIDIA. Last updated on May 10, 2024.