✨ Add a command to run ironic locally using podman

[dtantsur]

The new command run-local-ironic generates a YAML file with Kubernetes
manifests and then applies it using "podman kube play" with host
networking. This allows easy local testing of both IrSO and ironic-image
and can replace the existing bash script in BMO.

Closes: #86
Generated-By: Claude Code
Signed-off-by: Dmitry Tantsur [email protected]

[tuminoid]

I like the idea of using k8s manifests for local ironic, removes a lot of maintenance burden. Some quick nits and questions from quick look as requested.

[dtantsur]

I haven't checked this command with dnsmasq yet, I suspect it might need a bit more work. But in the meantime, I suggest we merge the base version and expand on it if necessary.

[tuminoid]

I'll give it a go next week

[dtantsur]

So, dnsmasq was not too bad, I only had to switch to using "sudo".

[tuminoid]

I tested it manually, both scenarios work nicely.

Same nits reappeared, tell Claude to be consistent :)

[tuminoid]

Suggested change

	while [[ $ATTEMPT -ne 30 ]]
	do
	while [[ "${ATTEMPT}" -ne 30 ]]; do

[dtantsur]

It's a bit too much of a nitpick: we know how ATTEMPT looks like :) Will fix anyway

[dtantsur]

Same nits reappeared, tell Claude to be consistent :)

The bash part is fully on me :D

[Rozzii]

I would like to discuss the leagal background of this PR on wednesday a bit further as I beleive this is the first PR marked as AI generated that is proposed for the project.

I am worried because of the potential miss match between Claude's provider Anthropics AI usage policies, the LinuxFoundations rules https://www.linuxfoundation.org/legal/generative-ai (point 1 or rule 1) and our project's license.

Anthropic states here https://www.anthropic.com/legal/consumer-terms

that only non commercial use of Anthropic's tools is allowed. While LF asks the projects to make sure the "generative AI tool do not place any contractual restrictions on how the tool’s output can be used", IMO according to LF standards there is a conflict between our project's Apache liccence and Anthropic's general licence that states that the tools (thus their output) can be used for non commercial use.

I know I am not a lawyer that is why I tried to base my reasoning on the guidelines of the LF, and it seems to me very clear cut conflict here. For me the only question is whether we follow LF guidelines or not because we are a CNCF project but AFAIK cncf has no guideline of it's own.

[Rozzii]

/hold
I would like to hold this a little bit because of my legal worries. Let's discuss this on community meeting, I will try to also find someone who is much more well versed about this topic than me.

[JanMkl]

I would like to discuss the leagal background of this PR on wednesday a bit further as I beleive this is the first PR marked as AI generated that is proposed for the project.

I am worried because of the potential miss match between Claude's provider Anthropics AI usage policies, the LinuxFoundations rules https://www.linuxfoundation.org/legal/generative-ai (point 1 or rule 1) and our project's license.

Anthropic states here https://www.anthropic.com/legal/consumer-terms that only non commercial use of Anthropic's tools is allowed. While LF asks the projects to make sure the "generative AI tool do not place any contractual restrictions on how the tool’s output can be used", IMO according to LF standards there is a conflict between our project's Apache liccence and Anthropic's general licence that states that the tools (thus their output) can be used for non commercial use.

I know I am not a lawyer that is why I tried to base my reasoning on the guidelines of the LF, and it seems to me very clear cut conflict here. For me the only question is whether we follow LF guidelines or not because we are a CNCF project but AFAIK cncf has no guideline of it's own.

@Rozzii and @dtantsur I took a bit more closer look at the Anthropic Terms and just to first say not a lawyer but some experience from the commercial contract parts. Here is how I read the agreement.

Anthropic’s Consumer Terms are broadly compatible with the Linux Foundation’s policy for general use, but some ambiguous provisions — particularly around “non-commercial use” and conditional compliance requirements — create areas of potential risk.

The “non-commercial use only” language appears in Section 11 (Disclaimer of warranties, limitations of liability, and indemnity). Read in context, it can reasonably be interpreted as a limitation on Anthropic’s liability rather than a blanket prohibition on commercial use of outputs. If that interpretation is correct, there is no conflict with the LF policy. This view aligns with Section 4 (Inputs, Outputs, and Materials), which emphasizes that outputs “may be inaccurate,” making it the user’s responsibility to validate results. In practice, that means the individual submitting a pull request based on generated output bears responsibility for ensuring its accuracy and suitability.

To minimize risk and remain aligned with LF policy, it is best to:

• Use only paid or standard accounts (not evaluation accounts, which are explicitly restricted to non-commercial use).
• Avoid violating Anthropic’s Acceptable Use Policy (AUP), since rights in outputs are contingent on compliance. Notably, the AUP restricts use of outputs for training competing AI models.
• Exercise caution if outputs are intended for training or further AI model development (likely not a concern for this project).
• Record and rely on the interpretation that “non-commercial use” refers to liability disclaimers rather than a usage prohibition.

That said, because the wording is somewhat ambiguous and open to interpretation, it would be helpful for @krook to review this and confirm whether the ambiguity falls within acceptable limits under the LF and CNCF AI policy.

[dtantsur]

Use only paid or standard accounts (not evaluation accounts, which are explicitly restricted to non-commercial use).

Yep, I used a paid (by my employer) account.

[elfosardo]

@Rozzii @JanMkl doesn't that clause refer exclusively to the Service API ?
https://support.claude.com/en/articles/8987200-can-i-use-the-claude-api-for-individual-use
meaning, in short, that I can't resell my access. let someone else use it, or use it for commercial use
for example, I can't build a tool or an app that use the Claude API on the background
anything that is built using the Claude API, but does not use them directly, is not covered by that

[dtantsur]

Btw, I got a confirmation internally that these terms apply to us: https://www.anthropic.com/legal/commercial-terms

[JanMkl]

Btw, I got a confirmation internally that these terms apply to us: https://www.anthropic.com/legal/commercial-terms

These Commercial Terms are more aligned with LF generative AI policy than the earlier linked consumer terms. They eliminate or mitigate most ambiguities and restraints. Key changes:

• Clearer ownership of outputs, with disclaimed rights from Anthropic.
• Indemnification protection for authorized use of outputs.
• Absence of a “non-commercial use only” clause constraining liability or usage.

[Rozzii]

Btw, I got a confirmation internally that these terms apply to us: https://www.anthropic.com/legal/commercial-terms

I have checked, this license looks compatible IMO especially because of this part:

So I am okay with it, would it be possible to mark specifically that the code was created by commercial Claude Code? Is there some sort of metadata that you could add @dtantsur for future reference in the git history?

[tuminoid]

Btw, I got a confirmation internally that these terms apply to us: https://www.anthropic.com/legal/commercial-terms

...

So I am okay with it, would it be possible to mark specifically that the code was created by commercial Claude Code? Is there some sort of metadata that you could add @dtantsur for future reference in the git history?

We discussed this in community meeting some time back. IIRC we agreed that we will follow the LF policy and the contributor is responsible for committed code, do we really nee to even mark anything about the use of AI? Everyone is using some sort of AI assist (code completion alone is almost 100% AI these days, and generated/assisted content is very integrated into the editor workflow), do we really want to even have it displayed? Is it RH policy to show it, or personal preference?

[dtantsur]

Is it RH policy to show it, or personal preference?

Since LF policy is basically "YOLO", I'm following the OpenInfra (org. governing OpenStack) policies: https://openinfra.org/legal/ai-policy. The goal is essentially openness about AI usage since some folks find it a sensitive issue.

would it be possible to mark specifically that the code was created by commercial Claude Code?

I think it's a dead end. As Tuomo rightfully point out, we don't verify that you, for instance, are using IDE in a way that is compliant with this IDE's license. In the end, it's on me to comply with the licensing of whatever tooling I'm using. This is why we add Sign-Off to our commits: to take legal responsibility for the contributions.

[Rozzii]

I just asked to clarify the type of model because you were marking it anyway but whatever I guess the PR discussion is good enough historical record.
/hold cancel

[Rozzii]

/hold cancel

[dtantsur]

Thank you for the discussion folks! Please review the patch itself when you have time.

[tuminoid]

/approve

with one final nit. I've tested this manually for both scenarios, and it works nicely.

[metal3-io-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: tuminoid

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [tuminoid]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[elfosardo]

/lgtm