Feature/slinky slurm hyperpod eks #804
Open
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This commit introduces a Bash script that automates the deployment of Slinky on an Amazon SageMaker HyperPod EKS cluster. The script streamlines the setup process with the following key features: - Cluster configuration generation based on a CloudFormation stack - Creation of an FSx for Lustre storage class - Installation of the AWS Load Balancer Controller - Setup of Slinky prerequisites (Prometheus, cert-manager, Slurm Operator) - Dynamic Slurm cluster configuration based on instance types - Creation and verification of FSx Persistent Volume Claims (PVCs) - Slurm cluster deployment using Helm - Configuration of a Network Load Balancer (NLB) for login node access The script includes error handling, progress tracking, and detailed logging to ensure a smooth deployment experience. It is designed to support both G5 and P5 instance types by dynamically allocating resources based on GPU count and EFA support. Note: This script has only been tested on G5 instance types. Dynamic resource allocation is handled by the dynamic_pods_allocation function, which is currently invoked within the set_slurm_values function. This dynamic behavior has not been fully tested. If you encounter issues with G5 instance deployments, consider commenting out the call to dynamic_pods_allocation in set_slurm_values. The script should still work correctly in this mode to function for P5 instance types, where a single pod is deployed per node. slinky automation tidy up added back a deleted file
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Issue #, if available:
Automate SageMaker HyperPod Slurm Cluster Deployment on EKS
This commit introduces a bash script that automates the deployment of slinky on Amazon a SageMaker HyperPod EKS cluster. Key features include:
The script includes error handling, progress tracking, and detailed logging to ensure a smooth deployment process. It's designed to work with both G5 and P5 instance types, dynamically allocating resources based on the GPU count and EFA support of the chosen instances.
However the it is only tested on deployment on the g5 instance types. The resource allocation modification based on the instance type is done on the dynamic_pods_allocation function ( this function is not test due to resource capacity unavailability). Currently this function call is made in this set_slurm_values function. If for g5 instance type it does not work please comment out this line and retry. The commented version should also work for instance type p5 where only one pod is deployed in one node.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.