DX-596-together-gpu-clusters: sync with mintlify-docs#1277

[zainhas]

Syncs the together-gpu-clusters skill with togethercomputer/mintlify-docs#1277 ([TCL-7655] docs: update GPU cluster targeted scale-down guidance, merged commit 9c02409).

Changed docs files that triggered this sync

• docs/gpu-clusters-management.mdx — restructured the Targeted scale-down section to document the new cluster-operator behavior deployed to prod.

Skill changes

• Updated skills/together-gpu-clusters/references/cluster-management.md, "Targeted Scale-down" section:
  • Added the annotation-based approach (kubectl annotate node <node_name> node.together.ai/delete-node-on-scale-down=true) as the preferred Kubernetes marking method.
  • Kept kubectl cordon as an alternative for cases where you also want to block new pods immediately.
  • Kept the Slurm scontrol ... State=drain command for Slurm clusters.
  • Noted the Together Cloud UI Node is cordon/draining checkpoint that customers should wait for before triggering scale-down.
  • Added the recommendation to scale down one node at a time so the operator can drain each cleanly.
  • Called out that annotated and cordoned nodes are prioritized for deletion above all others.

No changes to SKILL.md, scripts, or other references — the update only affects operational Kubernetes/Slurm guidance in the cluster-management reference.

Generated by the Sync Skills Cursor Automation. Please review before merging.

[broly-code-security-scanner]

Broly Security Scan

Note

✅ Clean scan
No vulnerabilities detected in this PR.

Note

Re-scan this PR anytime with /broly scan — useful after /broly undismiss, or to refresh findings without a new push.

Broly — SAST (zai-org/GLM-5.2) · Secrets · SCA · GH Actions (zizmor) · Containers · SBOM · Powered by Together AI