-
Notifications
You must be signed in to change notification settings - Fork 49
Release new version for Health Monitoring Agent (1.0.819.0_1.0.267.0)… #265
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
… with minor improvements and bug fixes. 1. New feature NVML API Check to detect hardware failure. Disabled Nvidia SMI query check 2. HMA will be able to detect File system read only error 3. For compatibility with AL2023, Non-NVIDIA devices will use a separate daemonset for deployment.
- key: node.kubernetes.io/instance-type | ||
operator: In | ||
values: | ||
- ml.inf2.xlarge |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why there is inf instance?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Previous HMA version do support inf instance, and it also didn't have nvidia drivier so will also be affected by the AL2023 change.
…ce launch (#249) * Release new version for Health Monitoring Agent (1.0.790.0_1.0.266.0) with minor improvements and bug fixes. (#254) * changelog version update (#256) Co-authored-by: Mohamed Zeidan <[email protected]> * Fix README documentation and broken anchor links (#252) **Description** - Updated README.md to fix broken internal navigation links, corrected SDK import paths, added proper syntax highlighting to code blocks. - Fixed training SDK imports, observability utils import path, and cluster management workflow examples. **Testing Done** - Verified all anchor links work correctly in table of contents and usage sections - Cross-referenced SDK imports against actual source code in src/sagemaker/hyperpod/ - Validated CLI commands match implementation in hyp_cli.py - Confirmed code examples use correct class names and method signatures * Small bug fix to print debug messages for inference logger (PySDK) (#246) * Draft of inference logger bug fix * Draft fix of inference logger for SDK * Revert adding --debug flag * Add debug parameter to failing unit tests * Fix create_from_dict to not have hardcoded debug flag * Add code-coverage workflow to GitHub workflows (#257) * Add code coverage workflow * Update artifact version to v4 * Fixed report upload * Simplified workflow using tox.ini * Make sure coverage is on right source files * Bug fix for 0 percent code coverage error * Bump version to 3.2.2 (#260) * Bump version to 3.2.2 **Description** Update package version from 3.2.1 to 3.2.2 in pyproject.toml and setup.py files. **Testing Done** Version bump only - no functional changes requiring additional testing. * Changelog update for v3.2.2 **Description** Added detaisl for Health Monitoring Agent updates to changelog **Testing Done** Production canary failure fixes validated. * Changelog update for v3.2.2 **Description** Updated the release date to represent the correct date. **Testing Done** No breaking changes. * Bump hyperpod-pytorch-job-template to v1.1.2 **Description** Update hyperpod-pytorch-job-template version from 1.1.1 to 1.1.2 and add changelog entry for node-count validation revert. **Testing Done** Version bump and changelog update - node-count validation revert functionality verified. * Update readme to include review guidelines (#261) * Update PR template * Update template * Update template format * Update format * Fix readme * Feature: Delete Cluster Command (#250) * delete cluster stack * delete cluster stack * removed unnecessary file * unit tests * more modular code * refactored modular code * sdk code added and improved modularity * cleanup * removed silent failure for sdk * fixed unit tests * integ tests * 2 integ happycase tests * changed test to use iam role instead of s3 bucket --------- Co-authored-by: Mohamed Zeidan <[email protected]> * Code Coverage for Integ Tests (#262) * Code Coverage for Integ Tests * Making sure target of coverage is correct * Removing duplicate implementation * Release new version for Health Monitoring Agent (1.0.819.0_1.0.267.0) with minor improvements and bug fixes. (#265) 1. New feature NVML API Check to detect hardware failure. Disabled Nvidia SMI query check 2. HMA will be able to detect File system read only error 3. For compatibility with AL2023, Non-NVIDIA devices will use a separate daemonset for deployment. * Removing duplicate cluster-creating integ test (#266) * Access entry fix (#267) * Fix Slurm failures from missing orchestration key (#268) * slurm-eks-helper-fix * Small fix to test to reflect new changes * small fix after resolving merge conflict --------- Co-authored-by: Xichao Wang <[email protected]> Co-authored-by: Mohamed Zeidan <[email protected]> Co-authored-by: Mohamed Zeidan <[email protected]> Co-authored-by: papriwal <[email protected]> Co-authored-by: aviruthen <[email protected]> Co-authored-by: Zhaoqi <[email protected]> Co-authored-by: jiayelamazon <[email protected]>
…ce launch (#249) * Release new version for Health Monitoring Agent (1.0.790.0_1.0.266.0) with minor improvements and bug fixes. (#254) * changelog version update (#256) Co-authored-by: Mohamed Zeidan <[email protected]> * Fix README documentation and broken anchor links (#252) **Description** - Updated README.md to fix broken internal navigation links, corrected SDK import paths, added proper syntax highlighting to code blocks. - Fixed training SDK imports, observability utils import path, and cluster management workflow examples. **Testing Done** - Verified all anchor links work correctly in table of contents and usage sections - Cross-referenced SDK imports against actual source code in src/sagemaker/hyperpod/ - Validated CLI commands match implementation in hyp_cli.py - Confirmed code examples use correct class names and method signatures * Small bug fix to print debug messages for inference logger (PySDK) (#246) * Draft of inference logger bug fix * Draft fix of inference logger for SDK * Revert adding --debug flag * Add debug parameter to failing unit tests * Fix create_from_dict to not have hardcoded debug flag * Add code-coverage workflow to GitHub workflows (#257) * Add code coverage workflow * Update artifact version to v4 * Fixed report upload * Simplified workflow using tox.ini * Make sure coverage is on right source files * Bug fix for 0 percent code coverage error * Bump version to 3.2.2 (#260) * Bump version to 3.2.2 **Description** Update package version from 3.2.1 to 3.2.2 in pyproject.toml and setup.py files. **Testing Done** Version bump only - no functional changes requiring additional testing. * Changelog update for v3.2.2 **Description** Added detaisl for Health Monitoring Agent updates to changelog **Testing Done** Production canary failure fixes validated. * Changelog update for v3.2.2 **Description** Updated the release date to represent the correct date. **Testing Done** No breaking changes. * Bump hyperpod-pytorch-job-template to v1.1.2 **Description** Update hyperpod-pytorch-job-template version from 1.1.1 to 1.1.2 and add changelog entry for node-count validation revert. **Testing Done** Version bump and changelog update - node-count validation revert functionality verified. * Update readme to include review guidelines (#261) * Update PR template * Update template * Update template format * Update format * Fix readme * Feature: Delete Cluster Command (#250) * delete cluster stack * delete cluster stack * removed unnecessary file * unit tests * more modular code * refactored modular code * sdk code added and improved modularity * cleanup * removed silent failure for sdk * fixed unit tests * integ tests * 2 integ happycase tests * changed test to use iam role instead of s3 bucket --------- Co-authored-by: Mohamed Zeidan <[email protected]> * Code Coverage for Integ Tests (#262) * Code Coverage for Integ Tests * Making sure target of coverage is correct * Removing duplicate implementation * Release new version for Health Monitoring Agent (1.0.819.0_1.0.267.0) with minor improvements and bug fixes. (#265) 1. New feature NVML API Check to detect hardware failure. Disabled Nvidia SMI query check 2. HMA will be able to detect File system read only error 3. For compatibility with AL2023, Non-NVIDIA devices will use a separate daemonset for deployment. * Removing duplicate cluster-creating integ test (#266) * Access entry fix (#267) * Fix Slurm failures from missing orchestration key (#268) * slurm-eks-helper-fix * Small fix to test to reflect new changes * small fix after resolving merge conflict --------- Co-authored-by: Xichao Wang <[email protected]> Co-authored-by: Mohamed Zeidan <[email protected]> Co-authored-by: Mohamed Zeidan <[email protected]> Co-authored-by: papriwal <[email protected]> Co-authored-by: aviruthen <[email protected]> Co-authored-by: Zhaoqi <[email protected]> Co-authored-by: jiayelamazon <[email protected]>
… with minor improvements and bug fixes.
What's changing and why?
Before/After UX
Before: NA
After: NA
How was this change tested?
Tested through node repair and fault detection through a newly created HyperPod EKS cluster on G5 and Trn1 instances.
Are unit tests added?
Yes
Are integration tests added?
No
Reviewer Guidelines
One of the following must be true: