Introduce metal-bmc to the mini-lab

[Gerrit91]

Description

Closes #328.

For now this is just playing around to find out the problems.

Depends on:

• Allow passing static hosts to BMC reporter. metal-bmc#99
• Generic IPMI fallback go-hal#94
• Generic inband IPMI for mini-lab. metal-hammer#193
• Metal bmc static hosts metal-roles#736

Used AI-Tools ✨

• None used for generation

[Gerrit91]

Currently go-hal errors out because there is no default IPMI implementation and it wants to access Redfish:

Aug 24 17:28:12 leaf01 docker[29125]: {"time":"2026-08-24T17:28:12.259603908+02:00","level":"ERROR","msg":"could not establish outband connection to device bmc","mac":"1a:8e:cc:99:a9:28","ip":"172.42.0.5","err":"unable to establish redfish connection for ip:172.42.0.5 user:ADMIN error:Get \"https://172.42.0.5/redfish/v1/\": dial tcp 172.42.0.5:443: connect: connection refused"}
Aug 24 17:28:12 leaf01 docker[29125]: {"time":"2026-08-24T17:28:12.259672969+02:00","level":"ERROR","msg":"could not enrich all ipmi details","error":"unable to establish redfish connection for ip:172.42.0.6 user:ADMIN error:Get \"https://172.42.0.6/redfish/v1/\": dial tcp 172.42.0.6:443: connect: connection refused"}

[majst01]

Currently go-hal errors out because there is no default IPMI implementation and it wants to access Redfish:

Aug 24 17:28:12 leaf01 docker[29125]: {"time":"2026-08-24T17:28:12.259603908+02:00","level":"ERROR","msg":"could not establish outband connection to device bmc","mac":"1a:8e:cc:99:a9:28","ip":"172.42.0.5","err":"unable to establish redfish connection for ip:172.42.0.5 user:ADMIN error:Get \"https://172.42.0.5/redfish/v1/\": dial tcp 172.42.0.5:443: connect: connection refused"}
Aug 24 17:28:12 leaf01 docker[29125]: {"time":"2026-08-24T17:28:12.259672969+02:00","level":"ERROR","msg":"could not enrich all ipmi details","error":"unable to establish redfish connection for ip:172.42.0.6 user:ADMIN error:Get \"https://172.42.0.6/redfish/v1/\": dial tcp 172.42.0.6:443: connect: connection refused"}

Maybe this could help: https://docs.openstack.org/sushy-tools/latest/user/dynamic-emulator.html and https://docs.airshipit.org/airshipctl/virtual_redfish_bmc.html

[Gerrit91]

Instead of using Redfish I used a generic IPMI fallback: metal-stack/go-hal#94

With this now the machines can be reported by the metal-bmc containing FRU info and more. Power commands also work.

Unfortunately, when the machine boots, the metal-hammer destroys the working IPMI connection details with some hard-coded data coming from the days back when we used vagrant: https://github.com/metal-stack/metal-hammer/blob/83a0aa531b5f6a24a87927cae2dbf82597cc1295/cmd/register/register.go#L339-L344. So, this should be looked at next.

❯ m machine ipmi
ID                                           POWER  IP               MAC                BOARD PART NUMBER  BIOS  BMC   SIZE          PARTITION  RACK       UPDATED  
00000000-0000-0000-0000-000000000001  ❗ ⭕  ⏻      172.42.0.6:6295  00:00:00:00:00:00  MINI-LAB-B1        0     9.08  v1-small-x86  mini-lab   test-rack  56s ago  
00000000-0000-0000-0000-000000000002  💀 ⭕  ⏾      172.42.0.5:6260  00:00:00:00:00:00  MINI-LAB-B1        0     9.08  v1-small-x86  mini-lab   test-rack  56s ago  

❯ m machine ipmi 00000000-0000-0000-0000-000000000001 -o json | jq .ipmi 
{
  "address": "172.42.0.6:6295",
  "bmcversion": "9.08",
  "fru": {
    "board_mfg": "metal-stack",
    "board_part_number": "MINI-LAB-B1",
    "chassis_part_number": "MINI-LAB-C12345",
    "chassis_part_serial": "MINI-LAB-CHASSIS-1",
    "product_manufacturer": "metal-stack",
    "product_part_number": "MINI-LAB-P1",
    "product_serial": "MINI-LAB-12345"
  },
  "interface": "lanplus",
  "last_updated": "2026-08-25T08:50:57.766Z",
  "mac": "00:00:00:00:00:00",
  "password": "vagrant",
  "powermetric": null,
  "powerstate": "ON",
  "powersupplies": null,
  "user": "vagrant"
}

[majst01]

Instead of using Redfish I used a generic IPMI fallback: metal-stack/go-hal#94

With this now the machines can be reported by the metal-bmc containing FRU info and more. Power commands also work.

Unfortunately, when the machine boots, the metal-hammer destroys the working IPMI connection details with some hard-coded data coming from the days back when we used vagrant: https://github.com/metal-stack/metal-hammer/blob/83a0aa531b5f6a24a87927cae2dbf82597cc1295/cmd/register/register.go#L339-L344. So, this should be looked at next.

❯ m machine ipmi
ID                                           POWER  IP               MAC                BOARD PART NUMBER  BIOS  BMC   SIZE          PARTITION  RACK       UPDATED  
00000000-0000-0000-0000-000000000001  ❗ ⭕  ⏻      172.42.0.6:6295  00:00:00:00:00:00  MINI-LAB-B1        0     9.08  v1-small-x86  mini-lab   test-rack  56s ago  
00000000-0000-0000-0000-000000000002  💀 ⭕  ⏾      172.42.0.5:6260  00:00:00:00:00:00  MINI-LAB-B1        0     9.08  v1-small-x86  mini-lab   test-rack  56s ago  

❯ m machine ipmi 00000000-0000-0000-0000-000000000001 -o json | jq .ipmi 
{
  "address": "172.42.0.6:6295",
  "bmcversion": "9.08",
  "fru": {
    "board_mfg": "metal-stack",
    "board_part_number": "MINI-LAB-B1",
    "chassis_part_number": "MINI-LAB-C12345",
    "chassis_part_serial": "MINI-LAB-CHASSIS-1",
    "product_manufacturer": "metal-stack",
    "product_part_number": "MINI-LAB-P1",
    "product_serial": "MINI-LAB-12345"
  },
  "interface": "lanplus",
  "last_updated": "2026-08-25T08:50:57.766Z",
  "mac": "00:00:00:00:00:00",
  "password": "vagrant",
  "powermetric": null,
  "powerstate": "ON",
  "powersupplies": null,
  "user": "vagrant"
}

All vagrant relicts should be removed.

[Gerrit91]

So, with an additional inband IPMI fallback and changes for vendor and board detection, it appears to work just fine. The metal-hammer kernel already has ipmi_sim modules loaded, so it can use a local IPMI device without requiring vagrant fake data anymore. It can also create the metal user.

I also tried to ipmitool sol activate against the created metal user, which works, too. So as a second step adding metal-console should also be able to work properly.