[devantler]

Why

KSail refuses to load the template's production cluster settings, so a new platform cannot create its production cluster and a platform that kept these settings cannot deploy to it. The settings let the cluster use every server it is allowed and leave none for the temporary server KSail needs while preparing node images.

What

The production cluster now tops out at one server fewer, keeping the same fixed nodes and allowing up to two extra nodes on demand instead of three, so KSail accepts the settings again. The autoscaling guide now gives the real node counts and explains why one server stays free.

Fixes #267

🤖 Generated with Claude Code

[devantler]

🤖 Generated by the Agentic Engineer

@codex review

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-25T02:21:59.228644Z	3c9983e	Manual request

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. 🚀

Reviewed commit: 3c9983e022

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[devantler]

🤖 Generated by the Agentic Engineer (interactive Claude Code session)

Ready for merge at 3c9983e. Review: Codex completed at this head with no findings ("Didn't find any major issues", reviewed commit 3c9983e022). CodeRabbit's included review for this repository was used up until about 02:36Z (stated on #266). 0 review threads.

• Tried as the platform owner would hit it: loaded ksail.prod.yaml through KSail v7.202.70's own validated configuration loader, the path ksail cluster create and ksail cluster update take, using a scratch harness inside a KSail checkout. It contacts no cluster.
  • main (control): fails, "autoscaler configuration leaves no Hetzner server slot for the Talos snapshot build: reachableTotal(10) … serverLimit(10)".
  • This head: loads cleanly. Control-plane and worker machine configs validate (22/16 documents), and the only remaining output is KSail's existing advisory about the install-image patch.
  • Current main (b13d8f0, with fix(talos): start new platforms on Kubernetes v1.36.4 and Talos v1.13.11 #266's new Talos/Kubernetes pins), alone: still fails with the same error. With this change applied: loads cleanly with Talos v1.13.11, kubelet v1.36.4 and the installer at v1.13.11. That is the config a new platform gets once both are merged.
• Why 9: 3 control planes and 4 static workers are kept. The pools still allow 3 autoscaled nodes, but the cluster total stops at 9, so one of the 10 default Hetzner servers stays free for KSail's temporary snapshot-build server. The field description in KSail (serverLimit: "Validation rejects configs whose reachable total … exceeds it") matches this reading.
• Docs: the autoscaling guide now gives the real counts (4 static workers; it said 3) and explains the reserved slot.
• Checks: all required checks pass. ksail workload validate for prod passes locally.
• Not verified: a real Hetzner create, because the template runs no cluster by design. Existing platforms own their ksail.prod.yaml, so they need the same one-line edit (noted on New production platforms fail KSail's config check: no server left for the snapshot build #267).