Redfish, IPMI & the BMC

Out-of-band server management for DGX and GPU clusters: what the BMC is, how IPMI and ipmitool work, how the Redfish REST API is organized, how to call its endpoints with curl and CLI tools, how IPMI and Redfish map to each other, and how BCM and NVIDIA tooling use them. Ends with four hands-on simulated tickets on a BCM head node. Objectives 1.3 and 5.2.

1. The big picture: in-band vs out-of-band

2. What the BMC is

• The Baseboard Management Controller is a small SoC with its own processor, firmware/OS, memory, and network port, soldered onto the server's motherboard.
• It runs on standby power: as long as the PSUs are plugged in, the BMC is alive, even with the host "off."
• It reads sensors (temperatures, fans, voltages, PSU status), controls power, logs hardware events in the SEL (System Event Log), stores FRU inventory (serial and part numbers), and offers remote KVM (keyboard/video/mouse) and virtual media (mount an ISO remotely).
• The host can also talk to the BMC locally through an internal interface (KCS), which is how ipmitool works on the host without a network (/dev/ipmi0).

3. IPMI: the classic protocol

4. ipmitool command tour

# Remote (out-of-band). Keep the password out of argv: -E reads IPMI_PASSWORD, or -f <file>
export IPMI_PASSWORD=<bmc-password>
ipmitool -I lanplus -H dgx01-bmc -U admin -E chassis power status

# Power
ipmitool ... chassis power status | on | off | cycle | reset | soft
#   off   = hard power off (like holding the button)
#   soft  = ACPI graceful shutdown request to the OS
#   cycle = off, then on       reset = hard reset (no power drop)

# Boot device (PXE for BCM reprovisioning)
ipmitool ... chassis bootdev pxe                          # next boot only
ipmitool ... chassis bootdev pxe options=persistent,efiboot

# Logs, sensors, inventory
ipmitool ... sel elist             # System Event Log, with decoded timestamps
ipmitool ... sel save dgx01-sel.txt   # keep a copy as evidence
ipmitool ... sel clear             # (save it first!)
ipmitool ... sdr elist             # all sensors with status
ipmitool ... sensor | grep -i temp
ipmitool ... fru print             # serial numbers, part numbers

# The BMC itself
ipmitool ... mc info               # BMC firmware version
ipmitool ... mc reset cold         # reboot the BMC (the host keeps running)
ipmitool ... lan print 1           # BMC IP, MAC, VLAN on channel 1
ipmitool ... user list 1

# Console and locating the server
ipmitool ... sol activate          # serial console over LAN (~. to exit)
ipmitool ... chassis identify 60   # blink the locator LED for 60 s

# Local (in-band), from the host OS. No credentials needed, requires root
sudo modprobe -a ipmi_devintf ipmi_si   # -a: load both (otherwise the 2nd word is a module parameter)
sudo ipmitool lan print 1          # "what is my BMC's IP?"

Which version, from where?

5. IPMI security weaknesses

# Is cipher suite 0 enabled? One character per suite, starting at suite 0:
#   X = disabled, a = up to Administrator, o = Operator, u = User
ipmitool ... lan print 1 | grep -i cipher
#   RMCP+ Cipher Suites   : 0,1,2,3,6,7,8,11,12,15,16,17
#   Cipher Suite Priv Max : aaaaXXaaaXXaaXX       <- first 'a' = suite 0 allowed as admin

ipmitool ... lan set 1 cipher_privs XaaaXXaaaXXaaXX   # first char X = suite 0 off

# Proof: suite 0 must now be refused, even with a wrong password
ipmitool -I lanplus -C 0 -H dgx05-bmc -U admin -P wrong chassis power status

6. Redfish: the modern REST API

$ curl -s https://dgx01-bmc/redfish/v1/ --cacert bmc-ca.pem | jq
{
  "@odata.id": "/redfish/v1/",
  "RedfishVersion": "1.15.0",
  "Systems":        { "@odata.id": "/redfish/v1/Systems" },
  "Chassis":        { "@odata.id": "/redfish/v1/Chassis" },
  "Managers":       { "@odata.id": "/redfish/v1/Managers" },
  "AccountService": { "@odata.id": "/redfish/v1/AccountService" },
  "SessionService": { "@odata.id": "/redfish/v1/SessionService" },
  "EventService":   { "@odata.id": "/redfish/v1/EventService" },
  "UpdateService":  { "@odata.id": "/redfish/v1/UpdateService" },
  "TaskService":    { "@odata.id": "/redfish/v1/TaskService" },
  ...
}

7. The Redfish resource tree

8. HTTP methods, status codes, and ETags

9. Authentication: Basic auth vs sessions

# Method 1: HTTP Basic auth on every request (simple, fine for one-off calls)
curl -s -u "$BMC_USER:$BMC_PASS" --cacert bmc-ca.pem https://$BMC/redfish/v1/Systems

# Method 2: Session login (better for scripts making many calls)
curl -si --cacert bmc-ca.pem -X POST https://$BMC/redfish/v1/SessionService/Sessions \
     -H 'Content-Type: application/json' \
     -d '{"UserName":"'"$BMC_USER"'","Password":"'"$BMC_PASS"'"}'
#   HTTP/1.1 201 Created
#   X-Auth-Token: 4f9c...            <- send this header on later requests
#   Location: /redfish/v1/SessionService/Sessions/17

curl -s -H "X-Auth-Token: $TOKEN" --cacert bmc-ca.pem https://$BMC/redfish/v1/Chassis

# Log out (frees a session slot on the BMC)
curl -s -X DELETE -H "X-Auth-Token: $TOKEN" --cacert bmc-ca.pem https://$BMC/redfish/v1/SessionService/Sessions/17

10. Common tasks with curl

R="curl -s -u $BMC_USER:$BMC_PASS --cacert bmc-ca.pem -H Content-Type:application/json"

# 1. Find the system ID (don't guess it)
$R https://$BMC/redfish/v1/Systems | jq -r '.Members[]."@odata.id"'
#   /redfish/v1/Systems/DGX

# 2. Power state and health
$R https://$BMC/redfish/v1/Systems/DGX | jq '{PowerState, Status}'
#   { "PowerState": "On", "Status": { "State": "Enabled", "Health": "OK" } }

# 3. Which reset types does this BMC allow?
$R https://$BMC/redfish/v1/Systems/DGX | jq '.Actions."#ComputerSystem.Reset"'
#   "ResetType@Redfish.AllowableValues": ["On","ForceOff","GracefulShutdown","GracefulRestart","ForceRestart","PowerCycle"]

# 4. Graceful reboot of the host
$R -X POST https://$BMC/redfish/v1/Systems/DGX/Actions/ComputerSystem.Reset \
   -d '{"ResetType":"GracefulRestart"}'

# 5. PXE boot once (for example to reprovision with BCM)
$R -X PATCH https://$BMC/redfish/v1/Systems/DGX \
   -d '{"Boot":{"BootSourceOverrideTarget":"Pxe","BootSourceOverrideEnabled":"Once"}}'

# 6. Temperatures and fans
$R https://$BMC/redfish/v1/Chassis/<chassis-id>/Thermal | jq '.Temperatures[] | {Name, ReadingCelsius, Status}'

# 7. Event log entries
$R https://$BMC/redfish/v1/Managers/<bmc-id>/LogServices | jq -r '.Members[]."@odata.id"'
$R https://$BMC/redfish/v1/Managers/<bmc-id>/LogServices/SEL/Entries | jq '.Members[] | {Created, Severity, Message}'

# 8. Reboot the BMC (not the host)
$R -X POST https://$BMC/redfish/v1/Managers/<bmc-id>/Actions/Manager.Reset -d '{"ResetType":"GracefulRestart"}'

More everyday tasks

# 9. Inventory
$R https://$BMC/redfish/v1/Systems/DGX | jq '{Model, SerialNumber, BiosVersion, ProcessorSummary, MemorySummary}'
$R https://$BMC/redfish/v1/Systems/DGX/Processors | jq -r '.Members[]."@odata.id"'
$R https://$BMC/redfish/v1/Systems/DGX/Memory     | jq '."Members@odata.count"'
$R https://$BMC/redfish/v1/Systems/DGX/Storage    | jq -r '.Members[]."@odata.id"'

# 10. Sensors (the newer model) and paging: big collections come back a page at a time
$R https://$BMC/redfish/v1/Chassis/DGX/Sensors | jq '."Members@odata.count", ."Members@odata.nextLink"'
$R "https://$BMC/redfish/v1/Chassis/DGX/Sensors?\$skip=75"
$R https://$BMC/redfish/v1/Chassis/DGX/Sensors/<sensor> | jq '{Name, Reading, ReadingUnits, Status}'

# 11. PSUs and power draw
$R https://$BMC/redfish/v1/Chassis/DGX/PowerSubsystem/PowerSupplies | jq -r '.Members[]."@odata.id"'
$R https://$BMC/redfish/v1/Chassis/DGX/Power | jq '.PowerControl[0].PowerConsumedWatts'   # older Power resource

# 12. Blink the identify LED so the datacenter tech pulls the right box
$R -X PATCH https://$BMC/redfish/v1/Chassis/DGX -d '{"LocationIndicatorActive":true}'
#   older schemas: {"IndicatorLED":"Blinking"}      ipmitool equivalent: chassis identify

# 13. BMC time (so SEL timestamps line up with host logs)
$R -X PATCH https://$BMC/redfish/v1/Managers/BMC/NetworkProtocol \
   -d '{"NTP":{"ProtocolEnabled":true,"NTPServers":["ntp1.example.internal"]}}'

# 14. Users: list, add a read-only monitoring account, change a password, check lockout policy
$R https://$BMC/redfish/v1/AccountService/Accounts | jq -r '.Members[]."@odata.id"'
$R -X POST https://$BMC/redfish/v1/AccountService/Accounts \
   -d '{"UserName":"monitor","Password":"'"$NEW_BMC_PASS"'","RoleId":"ReadOnly","Enabled":true}'
$R -X PATCH -H 'If-Match: *' https://$BMC/redfish/v1/AccountService/Accounts/<n> -d '{"Password":"'"$NEW_BMC_PASS"'"}'
$R https://$BMC/redfish/v1/AccountService | jq '{MinPasswordLength, AccountLockoutThreshold, AccountLockoutDuration}'

# 15. Save, then clear, the SEL
$R https://$BMC/redfish/v1/Managers/BMC/LogServices/SEL/Entries > sel-$(date +%F).json
$R -X POST https://$BMC/redfish/v1/Managers/BMC/LogServices/SEL/Actions/LogService.ClearLog -d '{}'

# 16. Query options: only if the service root advertises them
$R https://$BMC/redfish/v1/ | jq .ProtocolFeaturesSupported
$R "https://$BMC/redfish/v1/Systems/DGX?\$select=PowerState,Status"
$R "https://$BMC/redfish/v1/Systems?\$expand=.(\$levels=1)"

# 17. Replace the self-signed HTTPS certificate (so nobody needs curl -k)
$R -X POST https://$BMC/redfish/v1/CertificateService/Actions/CertificateService.GenerateCSR -d '{
  "CertificateCollection": {"@odata.id": "/redfish/v1/Managers/BMC/NetworkProtocol/HTTPS/Certificates"},
  "CommonName": "dgx01-bmc.example.internal", "Organization": "Example", "Country": "US"}' | jq -r .CSRString > bmc.csr
#   sign bmc.csr with your CA, then install it:
$R -X POST https://$BMC/redfish/v1/CertificateService/Actions/CertificateService.ReplaceCertificate -d '{
  "CertificateUri": {"@odata.id": "/redfish/v1/Managers/BMC/NetworkProtocol/HTTPS/Certificates/1"},
  "CertificateType": "PEM", "CertificateString": "-----BEGIN CERTIFICATE-----\n...\n-----END CERTIFICATE-----"}'

11. Firmware updates and tasks

# What firmware is installed?
$R https://$BMC/redfish/v1/UpdateService/FirmwareInventory | jq -r '.Members[]."@odata.id"'
$R https://$BMC/redfish/v1/UpdateService/FirmwareInventory/<item> | jq '{Name, Version, Updateable}'

# Push an update package (multipart, where supported)
curl -s -u "$BMC_USER:$BMC_PASS" --cacert bmc-ca.pem -X POST \
     "https://$BMC$(curl -s -u "$BMC_USER:$BMC_PASS" --cacert bmc-ca.pem https://$BMC/redfish/v1/UpdateService | jq -r .MultipartHttpPushUri)" \
     -F 'UpdateParameters={"Targets":[]};type=application/json' -F 'UpdateFile=@firmware.fwpkg'
#   UpdateParameters is a required part ({} or "Targets":[] if you have nothing to set).
#   Targets [] means "the service chooses"; on DGX H100 name the component, e.g.
#   {"Targets":["/redfish/v1/UpdateService/FirmwareInventory/HostBMC_0"]}
#   HTTP 202 Accepted. Location: the task monitor URI (e.g. /redfish/v1/TaskService/TaskMonitors/5);
#   the body is the Task (/redfish/v1/TaskService/Tasks/5).

# Poll the task until it completes
$R https://$BMC/redfish/v1/TaskService/Tasks/5 | jq '{TaskState, PercentComplete, Messages}'

# Pull variant: the BMC downloads the image itself (where UpdateService lists SimpleUpdate)
$R -X POST https://$BMC/redfish/v1/UpdateService/Actions/UpdateService.SimpleUpdate \
   -d '{"ImageURI":"https://repo.example.internal/fw/package.fwpkg","TransferProtocol":"HTTPS"}'
#   Push (MultipartHttpPushUri): you upload the file. Pull (SimpleUpdate): the BMC fetches a URL.

12. Events, logs, and telemetry

$R -X POST https://$BMC/redfish/v1/EventService/Subscriptions -d '{
  "Destination": "https://monitor.example.internal/redfish-events",
  "Protocol": "Redfish",
  "EventFormatType": "Event"
}'

13. CLI tools and libraries

14. IPMI ↔ Redfish mapping

15. DGX, HGX, and BCM specifics

• DGX BMC port: a dedicated RJ-45 management port that connects to the cluster's OOB management network, separate from the in-band management and compute fabrics.
• First-boot credentials: DGX systems require setting or changing the BMC password. Don't leave factory defaults on a reachable network.
• Remote KVM and virtual media: from the BMC web UI you can watch the console during boot and mount an ISO, for example to reinstall DGX OS.
• HGX baseboard management: on HGX/DGX systems, the GPU baseboard has its own management controller (the HMC, HGX Management Controller) that handles GPU and NVSwitch firmware and telemetry. It's typically reached through the BMC's Redfish service, which is why GPU-tray firmware appears in the Redfish firmware inventory. See the HGX firmware guide.
• In-band vs out-of-band on DGX: nvsm and DCGM run in the host OS (in-band). The BMC/Redfish works even when the OS is down.

BCM and the BMC

• BCM stores each node's BMC address and credentials (BMC settings at the cluster, category, or node level) and uses them for power control through IPMI or Redfish.
• That's how BCM can power-cycle a hung node or force a PXE reprovision without anyone touching the server.

cmsh
% device
% power status -n dgx01..dgx04
% power reset -n dgx03          # hard reset through the BMC
% power on -c dgx-h100          # a whole category

DGX H100/H200 Redfish recipes

These follow NVIDIA's DGX H100/H200 user guide ("Redfish APIs Support"). The DGX B200 and B300 guides document the same GPU power path; check your model's guide for the rest. $R is the curl alias from section 10.

# Reach the BMC from the host itself: the BMC sits at 169.254.0.17 on an internal link.
# Give the host's interface on that link (an enx... USB NIC) an address in 169.254.0.0/16:
sudo ip addr add 169.254.0.18/16 dev enx<...>            # NVIDIA's example uses ifconfig
nvfwupd -t ip=169.254.0.17 username=$BMC_USER password=$BMC_PASS show_version

# Set the BMC password after first boot: 13-20 characters, upper + lower + digit + special, no spaces
$R -X PATCH -H 'If-Match: *' https://$BMC/redfish/v1/AccountService/Accounts/2 -d '{"Password":"'"$NEW_BMC_PASS"'"}'

# Reboot the BMC / factory-reset the BMC configuration
$R -X POST https://$BMC/redfish/v1/Managers/BMC/Actions/Manager.Reset -d '{"ResetType":"ForceRestart"}'
$R -X POST https://$BMC/redfish/v1/Managers/BMC/Actions/Manager.ResetToDefaults -d '{"ResetType":"ResetAll"}'

# Firmware: push to UpdateService/upload; parameters.json picks the target
echo '{"Targets":["/redfish/v1/UpdateService/FirmwareInventory/HGX_0"]}' > parameters.json
curl -s -u "$BMC_USER:$BMC_PASS" --cacert bmc-ca.pem -H 'Expect:' -X POST https://$BMC/redfish/v1/UpdateService/upload \
     -F 'UpdateParameters=@parameters.json;type=application/json' -F 'UpdateFile=@nvfw_HGX_DGXH100-H200x8_<ver>.fwpkg'
#   GPU tray: always target HGX_0 (nvfw_HGX_... package).
#   CPU/motherboard tray: name the component, e.g. HostBMC_0 or CPLDMB_0 (nvfw_DGX_... package).
#   A package that isn't newer is skipped unless ForceUpdate is set (204 on success, 400 if already set):
$R -X PATCH -H 'If-Match: *' https://$BMC/redfish/v1/UpdateService -d '{"HttpPushUriOptions":{"ForceUpdate":true}}'
#   After the new image has booted, copy it to the other flash bank ("background copy"):
$R -X POST https://$BMC/redfish/v1/UpdateService/Actions/Oem/NvidiaUpdateService.CommitImage \
   -d '{"Targets":["/redfish/v1/UpdateService/FirmwareInventory/HostBIOS_0"]}'

# BIOS settings: find the attribute registry, read current values, PATCH the future settings (SD)
$R https://$BMC/redfish/v1/Registries | jq -r '.Members[]."@odata.id"' | grep BiosAttributeRegistry
$R https://$BMC/redfish/v1/Systems/DGX/Bios | jq .Attributes
$R -X PATCH -H 'If-Match: *' https://$BMC/redfish/v1/Systems/DGX/Bios/SD -d '{"Attributes":{"<AttrName>":"<Value>"}}'
#   Every BIOS attribute change needs a power cycle. A BIOS update afterwards needs another one.
$R -X POST https://$BMC/redfish/v1/UpdateService/Actions/Oem/NvidiaUpdateService.ClearNVRAM \
   -d '{"Targets":["/redfish/v1/UpdateService/FirmwareInventory/HostBIOS_0"]}'   # clear BIOS settings to defaults

# Persistent boot order (unlike BootSourceOverride "Once"): read it, decode the entries, PATCH SD
$R https://$BMC/redfish/v1/Systems/DGX/SD | jq .Boot.BootOrder              # ["Boot0000","Boot000F","Boot0004",...]
$R https://$BMC/redfish/v1/Systems/DGX/BootOptions/0004 | jq '{DisplayName, UefiDevicePath}'   # MAC is in the path
$R -X PATCH https://$BMC/redfish/v1/Systems/DGX/SD -H 'If-Match: "<etag from the GET>"' \
   -d '{"Boot":{"BootOrder":["Boot0004","Boot0000","Boot000F"]}}'

# Secure Boot state (replacing the Platform Key: disable, DELETE the PK cert, POST a new PEM, reboot)
$R https://$BMC/redfish/v1/Systems/DGX/SecureBoot | jq '{SecureBootEnable, SecureBootMode}'

# Virtual media: mount a DGX OS ISO over NFS (reinstall with nobody on site)
$R -X POST https://$BMC/redfish/v1/Managers/BMC/Actions/Oem/AMIVirtualMedia.EnableRMedia -d '{"RMediaState":"Enable"}'
$R https://$BMC/redfish/v1/Managers/BMC/VirtualMedia | jq -r '.Members[]."@odata.id"'          # CD1 .. CD4
$R -X POST https://$BMC/redfish/v1/Managers/BMC/VirtualMedia/CD1/Actions/VirtualMedia.InsertMedia \
   -d '{"Image":"//nfs.example.internal/isos/DGXOS-7.x.iso","TransferProtocolType":"NFS"}'       # returns a Task
$R https://$BMC/redfish/v1/Managers/BMC/VirtualMedia/CD1 | jq '{ImageName, Inserted, TransferProtocolType}'
#   Then put "UEFI: AMI Virtual CDROM0 1.00" first in the boot order and power-cycle.
#   Unmount afterwards: POST .../VirtualMedia/CD1/Actions/VirtualMedia.EjectMedia

# Sensors, GPU power, telemetry. DGX collections return 75 members per page.
$R https://$BMC/redfish/v1/Chassis/DGX/Sensors | jq -r '."Members@odata.nextLink"'       # .../Sensors?$skip=75
$R https://$BMC/redfish/v1/Managers/BMC/LogServices/SEL/Entries | jq '.Members | length'   # also paged at 75
$R https://$BMC/redfish/v1/Systems/HGX_Baseboard_0/Processors/GPU_SXM_1/EnvironmentMetrics | jq .PowerLimitWatts
#   Reading = current GPU power (W), SetPoint = current power limit, AllowableMin/AllowableMax = range
$R https://$BMC/redfish/v1/TelemetryService/MetricReportDefinitions/HGX_PlatformEnvironmentMetrics_0

# BMC debug bundle for NVIDIA support
$R -X POST https://$BMC/redfish/v1/Managers/BMC/LogServices/DiagnosticLog/Actions/LogService.CollectDiagnosticData \
   -d '{"DiagnosticDataType":"OEM","OEMDiagnosticDataType":"ALL"}'                    # returns a Task
$R https://$BMC/redfish/v1/TaskService/Tasks/<n> | jq '.TaskState, .Messages[].Message'   # names the Attachment
curl -s -u "$BMC_USER:$BMC_PASS" --cacert bmc-ca.pem -o nvidiadiag.tar.gz \
     https://$BMC/redfish/v1/Managers/BMC/LogServices/DiagnosticLog/Attachment/<file>.tar.gz

# BMC configuration backup and restore (upload an AES key first)
#   Managers/BMC/Actions/Oem/NvidiaManager.UploadAESKey, then .BackupConfig / .RestoreConfig
# Power capping: Managers/BMC/NodeManager (Domains, Policies)

16. Command builder

17. "Which tool?" chooser

18. Sample outputs

What the commands in this guide print. Expand a command to see a typical result and what to check in it.

Chassis Power is on

Set Boot Device to pxe

   1 | 11/02/2026 | 08:12:44 | Event Logging Disabled #0x07 | Log area reset/cleared | Asserted
   2 | 11/02/2026 | 08:15:03 | Power Unit #0x01 | Power off/down | Asserted
   3 | 11/02/2026 | 08:16:40 | System Event #0x52 | Timestamp Clock Sync | Asserted
   4 | 11/08/2026 | 02:14:51 | Power Supply PSU1_Status | Failure detected | Asserted
   5 | 11/08/2026 | 02:14:52 | Power Unit PSU_Redundancy | Redundancy Lost | Asserted
   6 | 11/08/2026 | 23:41:07 | OS Critical Stop #0x46 | Run-time critical stop | Asserted

Inlet_Temp       | 30h | ok  |  64.1 | 23 degrees C
CPU0_Temp        | 31h | ok  |   3.0 | 51 degrees C
CPU1_Temp        | 32h | ok  |   3.1 | 49 degrees C
GPU_SXM_1_Temp   | 40h | ok  |  11.1 | 38 degrees C
...
FAN_1_F          | 60h | ok  |  29.1 | 7800 RPM
FAN_2_F          | 61h | ok  |  29.2 | 7840 RPM
...
PSU0_Status      | 80h | ok  |  10.1 | Presence detected
PSU1_Status      | 81h | cr  |  10.2 | Presence detected, Failure detected
PSU0_Input_Power | 88h | ok  |  10.1 | 1020 Watts
...

FRU Device Description : Builtin FRU Device (ID 0)
 Chassis Type          : Rack Mount Chassis
 Chassis Part Number   : 965-24387-0002-003
 Chassis Serial        : 1660423000018
 Board Mfg Date        : Mon Mar  4 09:00:00 2024 UTC
 Board Mfg             : NVIDIA
 Board Product         : DGX H100
 Board Serial          : 1572423000101
 Board Part Number     : 699-24387-0001-000
 Product Manufacturer  : NVIDIA
 Product Name          : DGX H100
 Product Part Number   : 920-24387-2540-000
 Product Serial        : 1660423000018

FRU Device Description : PSU0 (ID 4)
 Product Manufacturer  : (example PSU vendor)
 Product Name          : 3300W CRPS
 Product Serial        : PS0-dgx01-4411

FRU Device Description : PSU1 (ID 5)
...

Device ID                 : 32
Device Revision           : 1
Firmware Revision         : 25.03
IPMI Version              : 2.0
Manufacturer ID           : 5703
Manufacturer Name         : NVIDIA
Product ID                : 5136 (0x1410)
Product Name              : Unknown (0x1410)
Device Available          : yes
Provides Device SDRs      : yes
Additional Device Support :
    Sensor Device
    SDR Repository Device
    SEL Device
    FRU Inventory Device
    IPMB Event Receiver
    IPMB Event Generator
    Chassis Device
Aux Firmware Rev Info     :
    0x01
    0x00
    0x00
    0x00

Set in Progress         : Set Complete
Auth Type Support       : MD5 PASSWORD
Auth Type Enable        : Callback : MD5 PASSWORD
                        : User     : MD5 PASSWORD
                        : Operator : MD5 PASSWORD
                        : Admin    : MD5 PASSWORD
                        : OEM      :
IP Address Source       : Static Address
IP Address              : 10.0.100.11
Subnet Mask             : 255.255.255.0
MAC Address             : 7c:c2:55:01:4a:1e
SNMP Community String   : AMI
IP Header               : TTL=0x40 Flags=0x40 Precedence=0x00 TOS=0x10
...
Default Gateway IP      : 10.0.100.1
Default Gateway MAC     : 00:00:5e:00:01:01
802.1q VLAN ID          : Disabled
802.1q VLAN Priority    : 0
RMCP+ Cipher Suites     : 0,1,2,3,6,7,8,11,12,15,16,17
Cipher Suite Priv Max   : XaaaXXaaaXXaaXX
                        :     X=Cipher Suite Unused
                        :     c=CALLBACK
                        :     u=USER
                        :     o=OPERATOR
                        :     a=ADMIN
                        :     O=OEM
Bad Password Threshold  : 0

Error in open session response message : no cipher suite match with proposed security algorithms

Error: Unable to establish IPMI v2 / RMCP+ session

[SOL Session operational.  Use ~? for help]

Ubuntu 22.04.4 LTS dgx01 ttyS0

dgx01 login:

HTTP/1.1 201 Created
Location: /redfish/v1/SessionService/Sessions/17
X-Auth-Token: token-redacted
OData-Version: 4.0
Content-Type: application/json; charset=utf-8
Cache-Control: no-store

{
  "@odata.id": "/redfish/v1/SessionService/Sessions/17",
  "@odata.type": "#Session.v1_5_0.Session",
  "Id": "17",
  "Name": "User Session",
  "UserName": "admin",
  "ClientOriginIPAddress": "10.0.100.254"
}

/redfish/v1/Systems/DGX

{
  "PowerState": "On",
  "Status": {
    "State": "Enabled",
    "Health": "OK"
  }
}

{
  "target": "/redfish/v1/Systems/DGX/Actions/ComputerSystem.Reset",
  "ResetType@Redfish.AllowableValues": [
    "On",
    "ForceOff",
    "GracefulShutdown",
    "GracefulRestart",
    "ForceRestart",
    "PowerCycle",
    "Nmi"
  ]
}

{
  "Name": "Inlet_Temp",
  "ReadingCelsius": 23,
  "Status": {
    "Health": "OK"
  }
}
{
  "Name": "CPU0_Temp",
  "ReadingCelsius": 51,
  "Status": {
    "Health": "OK"
  }
}
...

{
  "Created": "2026-11-08T02:14:51+00:00",
  "Severity": "Critical",
  "Message": "Power Supply: Failure detected"
}
{
  "Created": "2026-11-08T02:14:52+00:00",
  "Severity": "Critical",
  "Message": "Power Unit: Redundancy Lost"
}
{
  "Created": "2026-11-08T23:41:07+00:00",
  "Severity": "Critical",
  "Message": "OS Critical Stop: Run-time critical stop"
}
...

{
  "Model": "DGX H100",
  "SerialNumber": "1660423000018",
  "BiosVersion": "01.05.04",
  "ProcessorSummary": {
    "Count": 2,
    "Model": "Intel(R) Xeon(R) Platinum 8480C"
  },
  "MemorySummary": {
    "TotalSystemMemoryGiB": 2048
  }
}

212
"/redfish/v1/Chassis/DGX/Sensors?$skip=75"

{
  "TaskState": "Running",
  "PercentComplete": 45,
  "Messages": [
    {
      "@odata.type": "#Message.v1_1_1.Message",
      "Message": "The task with Id '5' has started.",
      "MessageArgs": [
        "5"
      ],
      "MessageId": "TaskEvent.1.0.TaskStarted",
      "Resolution": "None.",
      "Severity": "OK"
    }
  ]
}

[
  "Boot0000",
  "Boot000F",
  "Boot0004",
  "Boot0001",
  "Boot0002"
]

{
  "AllowableMax": 700,
  "AllowableMin": 200,
  "DefaultSetPoint": 700,
  "Reading": 312.47,
  "SetPoint": 700
}

19. Hands-on lab: simulated tickets

Five tickets from a simulated BCM head node, head01, which sits on the OOB management network with five DGX H100 nodes (dgx01–dgx05, BMCs dgx01-bmc–dgx05-bmc):

• OOB-101: a hung node. Read the SEL, keep the evidence, pick a power action that actually works on a hung OS, then find the GPU error in the host's log.
• RF-202: PXE-boot a node once with raw Redfish calls: service root, sessions, discovering IDs, ETags and If-Match, the Reset action, and logging out.
• FW-303: a BMC firmware update through UpdateService and TaskService on a BMC whose session slots are full, without touching the host's running job.
• NET-404: a new node's BMC has no usable address and allows IPMI cipher suite 0. Fix it in-band from the host, then prove BCM can power it and suite 0 is refused.
• HW-505: a failed PSU on a node that's running jobs. Collect the RMA evidence with ipmitool (sdr, sel, fru print, mc info), find the one fact IPMI can't give you, and light the locator LED, all without a power action.

20. Scenarios

21. Searchable reference

22. Flashcards (click to flip)

23. Exam traps in one list

24. Practice quiz (43 questions)