[feature] Ability to Add BMC to Machine Config for Existing Talos Nodes

[jd4883]

Problem Description

A feature that was part of sidero metal and apparently is still available (although maybe not well documented in the setup guides for talos/omni) is the ability to configure a BMC for servers that support it.

Based on documentation provided by @frezbo via slack, the feature is available as part of the bare metal infrastructure provider. The documentation indicates onboarding a server to this setup will wipe all disks.

Solution

It would be really nice to be able to enable BMC controls for already onboarded servers in an omni cluster. I have 4 nodes which all have BMC integrations available. It would be really nice to have a single management plane (omni) to start, shutdown, and reboot my machines rather than have to do so on the physical server or directly with each BMC interface.

That said, only 1 of my nodes could be safely wiped without major data loss. I'd rather be able to enable this on all 4 nodes. As a user, I would like to be able to add a BMC configuration (maybe part of the machine config) that allows credentials to be provided to my talos machines without requiring any disk wiping.

Based on documentation this is not a feature at this time.

Alternative Solutions

The alternative right now is to leave the BMC configuration out of the equation. I only have 4 nodes so maybe I am getting less value than someone with a bigger cluster, but I can see a lot of potential value of not having to navigate to each BMC to start servers up. I understand this is a convenience feature I am asking for but it would be really helpful.

Notes

No response

[smira]

It's not quite clear what exactly you're asking about. Are you using Omni's bare-metal infrastructure provider today, or not?

If you set up your machines manually without bare-metal provider, you're asking about adopting your machines into the provider, which is not available right now. (And it was never available in Sidero Metal).

[jd4883]

@smira let me clarify, I'm not trying to imply there is a missing feature from sidero metal but comparing the user experience of adding a machine.

Before adopting talos/omni, I fiddled a bit with sidero metal and (this may be because of using the bare metal infra provider) was able to just add a block of BMC configuration which worked. As someone who owns a few servers that have BMC capabilities, my hope would be as a user to be able to do a few things that aren't an option right now:

1. Take a manually provisioned node from an ISO or other deployment method and onboard it to the bare metal provider (honestly, this is probably what I should have set up with when I created my cluster but didn't know better), which would allow adding a BMC configuration.
2. As a user, I would like to be able to do operations like this with a choice of whether to wipe my node (a prompt in the UI or a warning before proceeding with a confirmation from the CLI would do the trick). This is assuming the BMC configuration cannot be implemented on nodes onboarded from an ISO that I am not wanting to wipe.

A simpler way to look at what I am hoping to achieve is having a way to add a BMC configuration to my worker/control plane nodes without being required to wipe them.

Also please clarify for my understanding - if hypothetically I were to remake my cluster from scratch and use the bare metal provider, for each of the conditions below, would my nodes be expected to wipe all disks or simply the disk where talos is installed?

1. Onboarding a node to the infrastructure provider which was previously installed via ISO
2. Enabling a BMC configuration on a node and accepting it into the cluster (my understanding is this part wipes all the nodes disks)
3. Removing a node from a cluster that has been accepted from the infrastructure provider
4. Other operations that as a user I should be extra careful about due to risk of wiping a node.

The idea behind this feature request is more user agency with better node level observability and hopefully power controls. If this isn't in the works or a priority that's not a big deal, but it would be really nice to have more flexibility here as a user.

If any of my understanding from what I have described is factually wrong or does not make sense please clarify.

Thanks in advance, this isn't an urgent request but would be nice to see if feasible.

Cheers,

Jacob