Revamp GKE A3 Mega blueprint and align integration tests#5483
Conversation
Summary of ChangesHello, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed! This pull request refactors the GKE A3 Mega blueprint to enhance deployment stability and predictability. By transitioning to locally version-controlled GPUDirect manifests and refining the integration test suite, the changes ensure a more robust and customizable environment for A3 accelerator instances. Highlights
Using Gemini Code AssistThe full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips. Invoking Gemini You can request assistance from Gemini at any point by creating a comment using either
Customization To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a Limitations & Feedback Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for GitHub and other Google products, sign up here. Footnotes
|
There was a problem hiding this comment.
Code Review
This pull request introduces a comprehensive example for deploying A3 Mega GKE clusters, featuring detailed documentation, infrastructure blueprints, and manifests for GPU performance testing. Key additions include the NCCL TCPXO installer and NRI device injector manifests, alongside updates to the validation Ansible playbook to ensure robust testing. Review feedback highlights several opportunities to improve maintainability and portability: specifically, addressing a 'bash-ism' in the installer script that could fail in non-bash environments, removing redundant sudo commands, and converting static manifests into templates to avoid hardcoding accelerator types and software versions.
2ee5a53
into
GoogleCloudPlatform:develop
Summary
This PR natively bundles GPUDirect manifests and refactors the A3 Mega blueprint for improved stability, predictability, and standardization across A* accelerator instances.
Motivation
Changes
examples/gke-a3-megagpu/):nccl-tcpxo-installer.yaml.tftpl,nri-device-injector.yaml, andnccl-test-latest.yamlgke-a3-megagpu.yaml): Updated workload component install step to apply GPUDirect manifests directly during deployment; setinstall_gpu_direct_manifests: falseon node poolTesting
A3M Spot Integration Tests: Validated with improved test automation including explicit DaemonSet verification and pod readiness checks.
Submission Checklist
NOTE: Community submissions can take up to 2 weeks to be reviewed.
Please take the following actions before submitting this pull request.