fix: Update a3m and a3u script to resolve slurm nccl test failure#5308
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 addresses an issue with Highlights
Changelog
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. You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension. Footnotes
|
There was a problem hiding this comment.
Code Review
This pull request adds a workaround for enroot in non-interactive shells where XDG_RUNTIME_DIR is not set. However, it introduces a critical security vulnerability in the import_pytorch_container.sh scripts. The scripts set up a fallback XDG_RUNTIME_DIR in /tmp using a predictable path and insecure permissions (sudo chmod a+rw), which is highly susceptible to symlink attacks leading to local privilege escalation to root. Furthermore, the test-nccl.yml configuration was modified to use a personal fork and branch, which must be reverted. The new logic is also duplicated across two separate scripts, and it would be better to centralize it in a shared script to improve maintainability.
examples/machine-learning/a3-megagpu-8g/nccl-tests/import_pytorch_container.sh
Outdated
Show resolved
Hide resolved
examples/machine-learning/a3-ultragpu-8g/nccl-tests/import_pytorch_container.sh
Outdated
Show resolved
Hide resolved
tools/cloud-build/daily-tests/ansible_playbooks/test-validation/test-nccl.yml
Outdated
Show resolved
Hide resolved
tools/cloud-build/daily-tests/ansible_playbooks/test-validation/test-nccl.yml
Outdated
Show resolved
Hide resolved
|
The PR tests have passed for A3M and A3U: |
|
PR tests for a3m and a3u are successful after modifications to the commands: |
|
We need to modify the git clone path in test-nccl.yml from main to current branch in order to test out our changes. So, reverted those changes in order to finalise the release to main. Ideally the recent working branch changes should be taken into account. That effort is being tracked under a separate ticket. |
8eb5de0
into
GoogleCloudPlatform:release-candidate
This PR resolves the error encountered while running NCCL test for a3 machines:
Error encountered in the NCCL tests:
mkdir: cannot create directory ‘/run/enroot’: Permission deniedFix:
XDG_RUNTIME_DIRis a standard Linux environment variable defined by the XDG Base Directory Specification. It points to a directory specific to the logged-in user for storing small, temporary runtime files. Enroot uses this variable to determine where to create its workspace. If it's unset, Enroot falls back to a hardcoded system path (/run) which causes the crash.We are resolving the Permission denied error by manually setting the
XDG_RUNTIME_DIRenvironment variable to a user-specific, world-writable directory in /tmp (e.g., /tmp/enroot-runtime-$(id -u)). This ensures enroot uses a safe, writable workspace instead of defaulting to the restricted, root-owned /run/enroot pathSubmission Checklist
NOTE: Community submissions can take up to 2 weeks to be reviewed.
Please take the following actions before submitting this pull request.