cray-hpe/gru

A utility for reading and modifying BMCs (e.g. iLO, RMMC) using RedFish (gofish).

Go Redfish Utility (gru)

Table of Contents

• Usage
• Development

An asynchronous RedFish utility for interrogating and managing blade management controllers.

Usage

gru help

Setting Credentials

• Environment Variables

  export USERNAME=admin
  export PASSWORD=password
  gru show system myserver-bmc.local

• Configuration File

  ---
  username: admin
  password: password

• Configuration file with a general username/password and exceptions (for server10 use a different username and password)

  ---
  username: admin
  password: password
  hosts:
    server10:
        username: root
        password: password

Querying Servers

• Pass any number of hosts, delimited by spaces.

  gru show system myserver-bmc.local myotherserver-bmc.local

• Piping a list a space delimited list of hosts.

  grep -oP pattern /etc/hosts | tr -s '\n' ' ' | gru show system

Development

make gru
./gru

gru's Issues

`gru bios set --clear-cmos` does not work on intel

SUMMARY

Potentially related to stmcginnis/gofish#268 (fixed in gofish 0.15.0)

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Run gru bios set --clear-cmos
2. Cold boot the node

EXPECTED RESULTS

BIOS is cleared.

ACTUAL RESULTS

BIOS remains the same, error is observed:

unable to execute request, no target provided

Printing BIOS Attributes is unsorted

SUMMARY

get bios (a.k.a. bios get) is not sorted, making it difficult to read.

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. gru bios get <node>

EXPECTED RESULTS

Output is sorted.

ACTUAL RESULTS

Output is only sorted for --json.

./gru --insecure bios get redbull-ncn-m001-mgmt.hpc.amslabs.hpecorp.net
Asynchronously querying [    1] hosts ... 
redbull-ncn-m001-mgmt.hpc.amslabs.hpecorp.net:
        CbRelaxedOrdering                       : 0                                                           
        PcieOpROMCtrl_4                         : 1                                                           
        PcieOpROMCtrl_9                         : 1                                                           
        NtbSBar01Prefetch_0                     : 0                                                           
        ProcessorSmxEnable                      : 0                                                           
        RdtCatOpportunisticTuning               : 0                                                           
        PciePortLinkSpeed_S_66                  : 3                                                           
        PciePortLinkSpeed_S_65                  : 3                                                           
        PciePortDisable_37                      : 2                                                           
        UFSDisable                              : 0                                                           
        VmdEnable_2                             : 0                 
// Truncated ...

`gru chassis boot <something> --now` does not work if the node is off

SUMMARY

If a node is off, the --now flag for boot does not work.
The flag does work if a node is already on, the flag triggers a reset as expected.

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Power off a node
2. Run gru chassis boot bios --now <node>
3. Node remains off

EXPECTED RESULTS

Node should power up.

ACTUAL RESULTS

Node remains off.

Invalid command order `bios get` vs `get bios` and duplicate code

SUMMARY

During the development of #18 , the app definition had an amendment to make development of the bios command easier.

Currently pkg.cli.bios.get and pkg.cli.bios.set have duplicate code, the original get bios and set bios structure of the CLi made implementing non-duplicate code difficult.

The new format, bios get and bios set allow the duplicate code to inherently get slurped/consolidated into bios.go.

Switching to the new command line syntax will clean up the duplicate code.

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

EXPECTED RESULTS

• gru bios get
• gru bios set

ACTUAL RESULTS

• gru get bios
• gru set bios

`gru --version` reports "dirty" on 0.0.4

SUMMARY

Official releases of gru show -dirty suffix when running gru --version.

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Download gru 0.0.4 https://github.com/Cray-HPE/gru/releases/tag/v0.0.4
2. Run gru --version

EXPECTED RESULTS

Any copy of gru only shows -dirty when built from a repository with uncommitted changes.

ACTUAL RESULTS

gru from the 0.0.4 release tag shows -dirty.

`gru --json chassis boot <something> --now` does not return a valid JSON object

SUMMARY

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Run gru --json chassis boot <something> --now

EXPECTED RESULTS

Printed value should be a parseable JSON object, instead of two separate JSON objects.

ACTUAL RESULTS

JSON objects should be merged, so jq or any other JSON interpreter may parse the output.

redbull-management:~/doomslayer # gru -j chassis boot bios --now ncn-s001-mgmt ncn-s002-mgmt ncn-s003-mgmt
{
  "ncn-s001-mgmt": {
    "target": "BiosSetup"
  },
  "ncn-s002-mgmt": {
    "target": "BiosSetup"
  },
  "ncn-s003-mgmt": {
    "target": "BiosSetup"
  }
}
{
  "ncn-s001-mgmt": {
    "previousPowerState": "On",
    "resetType": "ForceRestart"
  },
  "ncn-s002-mgmt": {
    "previousPowerState": "On",
    "resetType": "ForceRestart"
  },
  "ncn-s003-mgmt": {
    "previousPowerState": "On",
    "resetType": "ForceRestart"
  }
}

boot commands fail on HPE/iLO

SUMMARY

When invoking either of the following against an iLO machine, 404 errors are returned.

• gru show boot
• gru chassis boot

Found against the following node

{
  "ncn-m003-mgmt": {
    "biosVersion": "A43 v2.68 (02/06/2023)",
    "firmwareVersion": "iLO 5 v2.81",
    "processorModel": "AMD EPYC 7302P 16-Core Processor               ",
    "manufacturer": "HPE",
    "model": "ProLiant DL325 Gen10 Plus"
  }
}

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Find an iLO machine
2. Run gru show boot

EXPECTED RESULTS

A list of boot options should print out

intel-mgmt-node:
	Order:
		CRAY UEFI OS 1
		CRAY UEFI OS 0
		UEFI Samsung Flash Drive 1100
		UEFI HTTPv6: Network 00 at Riser 03 Slot 01
		UEFI IPv4: Network 00 at Riser 03 Slot 01
		UEFI HTTPv4: Network 00 at Riser 03 Slot 01
		UEFI IPv6: Network 00 at Riser 03 Slot 01
		UEFI HTTPv6: Network 01 at Riser 03 Slot 01
		UEFI IPv4: Network 01 at Riser 03 Slot 01
		UEFI HTTPv4: Network 01 at Riser 03 Slot 01
		UEFI IPv6: Network 01 at Riser 03 Slot 01
		UEFI HTTPv6: Network 00 at Riser 03 Slot 02
		UEFI IPv4: Network 00 at Riser 03 Slot 02
		UEFI HTTPv4: Network 00 at Riser 03 Slot 02
		UEFI IPv6: Network 00 at Riser 03 Slot 02
		UEFI HTTPv6: Network 01 at Riser 03 Slot 02
		UEFI IPv4: Network 01 at Riser 03 Slot 02
		UEFI HTTPv4: Network 01 at Riser 03 Slot 02
		UEFI IPv6: Network 01 at Riser 03 Slot 02
		UEFI HTTPv6: Intel Network 00 at Baseboard
		UEFI IPv4: Intel Network 00 at Baseboard
		UEFI HTTPv4: Intel Network 00 at Baseboard
		UEFI IPv6: Intel Network 00 at Baseboard
		UEFI HTTPv6: Intel Network 01 at Baseboard
		UEFI IPv4: Intel Network 01 at Baseboard
		UEFI HTTPv4: Intel Network 01 at Baseboard
		UEFI IPv6: Intel Network 01 at Baseboard
		UEFI SAMSUNG MZ7LH480HAHQ-00005 S45PNA0M839302
		Launch EFI Shell
		Enter Setup
		Boot Device List
		Network Boot
		UEFI SAMSUNG MZ7LH480HAHQ-00005 S45PNA0M839290

ACTUAL RESULTS

A 404 error is returned.

Overloaded `types.go` and `MapPrint`

SUMMARY

ISSUE TYPE

• Bug Report

STEPS TO REPRODUCE

1. Types specific to subpackages are living in a common area
2. cli.MapPrint is now coupled with various types from various subpackages

EXPECTED RESULTS

All subpackages should be modular.

1. cli.MapPrint should not be coupled to a subpackage's types.
2. pkg.cmd.clie.bios.amd.* should only be used when amd or epyc processors are detected.

ACTUAL RESULTS

Some subpackages are no longer modular, and context is lost.

1. cli.MapPrint now has special handling for a handful of types, instead Polymorphesim or a more generic handling should be used.
2. Some specific code, such as pkg.cmd.cli.bios.amd.epyc.rome is used in liberally in every single call.

Include CPU architecture in system print out (for every CPU)

SUMMARY

Include a list of every physical processor on the server when running show system:

• Architecture
• Model
• The socket index
• Core count

This would replace the ProcessorModel read from ProcessorSummary.Model.

ISSUE TYPE

• RFE

ADDITIONAL INFORMATION

I briefly tried implementing this, and while it works for JSON it isn't working for pretty-print.
Regardless, here's the formatting I was thinking:

gru show system redbull-ncn-m001-mgmt.hpc.amslabs.hpecorp.net drax-ncn-m001-mgmt.hpc.amslabs.hpecorp.net surtur-ncn-m001-mgmt.hpc.amslabs.hpecorp.net -j

{
  "drax-ncn-m001-mgmt.hpc.amslabs.hpecorp.net": {
    "biosVersion": "C38",
    "firmwareVersion": "12.84.17",
    "processors": [
      {
        "architecture": "x86",
        "cores": 24,
        "model": "AMD EPYC 7402 24-Core Processor",
        "socket": "P0"
      }
    ],
    "manufacturer": "Cray Inc.",
    "model": "R272-Z30-00"
  },
  "redbull-ncn-m001-mgmt.hpc.amslabs.hpecorp.net": {
    "biosVersion": "SE5C620.86B.02.01.0016.C0001.032120230338",
    "firmwareVersion": "2.88.71773d70",
    "processors": [
      {
        "architecture": "x86",
        "cores": 28,
        "model": "Intel Xeon processor",
        "socket": "CPU 1"
      },
      {
        "architecture": "x86",
        "cores": 28,
        "model": "Intel Xeon processor",
        "socket": "CPU 2"
      }
    ],
    "manufacturer": "Intel Corporation",
    "model": "S2600WFT"
  },
  "surtur-ncn-m001-mgmt.hpc.amslabs.hpecorp.net": {
    "biosVersion": "A43 v1.38 (10/30/2020)",
    "firmwareVersion": "iLO 5 v2.44",
    "processors": [
      {
        "architecture": "x86",
        "cores": 16,
        "model": "AMD EPYC 7302P 16-Core Processor               ",
        "socket": "Proc 1"
      }
    ],
    "manufacturer": "HPE",
    "model": "ProLiant DL325 Gen10 Plus"
  }
}

Set user credentials

SUMMARY

In recent discussions, IPMI was used often for this, but a more vendor agnostic approach is needed during hardware inventory/discovery as well as ad-hoc usages.

ISSUE TYPE

• RFE

ADDITIONAL INFORMATION

Perhaps this could also include creating and deleting users.