Skip to content

Add GPU scheduling and Karpenter examples - #26

Merged
RayLi-Muye merged 1 commit into
mainfrom
feature/gpu-scheduling-karpenter-examples
Jun 19, 2026
Merged

RayLi-Muye merged 1 commit into
mainfrom
feature/gpu-scheduling-karpenter-examples

Conversation

@RayLi-Muye

Copy link
Copy Markdown
Owner

Closes #2.

Scope

  • Adds CPU postprocess scheduling fields to the chart values, schema, and Job template.
  • Expands the AWS GPU/SQS example with GPU worker and CPU postprocess scheduling profiles.
  • Documents the Karpenter/EKS Auto Mode boundary without creating node pools or cloud resources.
  • Updates README, architecture notes, evidence map, delivery prompts, and chart contract tests.

Acceptance Criteria

  • GPU worker example includes nodeSelector, tolerations, and nvidia.com/gpu limits.
  • CPU postprocess example includes nodeSelector, tolerations, and affinity hints.
  • Docs explain why API-facing pods should stay off GPU nodes.
  • Docs describe Karpenter/EKS Auto Mode as external integration points, not resources provisioned by this chart.

Validation

  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make test
  • make smoke-local
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make lint
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template >/tmp/renderfarm-template-aws.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template-local >/tmp/renderfarm-template-local.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template-minio >/tmp/renderfarm-template-minio.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH scripts/render-examples.sh
  • python3 -m json.tool charts/render-worker/values.schema.json >/tmp/render-worker-values.schema.json
  • git diff --check
  • python3 /Users/winnie/.codex/skills/autoreview/scripts/autoreview --mode local

Docs Impact

Adds docs/scheduling.md and links it from README. Updates architecture and evidence docs for GPU worker, CPU postprocess, and API scheduling boundaries.

Risk And Rollback

Risk is low: the new postprocess scheduling defaults are empty and existing values continue rendering the same resources unless users opt in. Roll back by reverting this PR if scheduling examples or schema additions need to be revised.

Boundary

This PR does not deploy, release, provision Kubernetes nodes, create Karpenter NodePools, create EKS Auto Mode pools, access cloud credentials, or touch production data.

@RayLi-Muye
RayLi-Muye merged commit 0657da6 into main Jun 19, 2026
1 check passed
@RayLi-Muye
RayLi-Muye deleted the feature/gpu-scheduling-karpenter-examples branch June 19, 2026 01:20
@RayLi-Muye

Copy link
Copy Markdown
Owner Author

Post-merge validation summary:

Local validation before merge:

  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make test
  • make smoke-local
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make lint
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template >/tmp/renderfarm-template-aws.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template-local >/tmp/renderfarm-template-local.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH make template-minio >/tmp/renderfarm-template-minio.yaml
  • PATH=/tmp/codex-renderfarm-tools/bin:$PATH scripts/render-examples.sh
  • python3 -m json.tool charts/render-worker/values.schema.json >/tmp/render-worker-values.schema.json
  • git diff --check
  • python3 /Users/winnie/.codex/skills/autoreview/scripts/autoreview --mode local

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

Add GPU node scheduling and Karpenter examples

1 participant