fix(ray.sub): revalidate log dir before sentinel existence checks - #4004
Open
pulkitkumar95 wants to merge 1 commit into
Open
pulkitkumar95 wants to merge 1 commit into
pulkitkumar95 wants to merge 1 commit into
Conversation
Workers poll [[ -f STARTED_RAY_HEAD ]] on the shared filesystem; a stale negative dentry can hide the file for many minutes after the head creates it. Observed on a 32-node run: 52/128 worker units joined while the file existed for 8+ minutes, and the startup deadline killed three consecutive job windows. readdir on the log dir forces revalidation before the checks.
Author
|
/ok to test c080856 |
svcnvidia-nemo-ci
added
the
waiting-on-maintainers
This branch has not been deployed
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
What does this PR do ?
Prevents multi-node job startup failures caused by workers reading a stale negative dentry for the
STARTED_RAY_HEADsentinel on shared filesystems.Issues
Workers poll
[[ -f $LOG_DIR/STARTED_RAY_HEAD ]]; a stale negative dentry can hide the file for many minutes after the head creates it. Observed on a 32-node run: only 52/128 worker units joined while the sentinel had existed for 8+ minutes, and the startup deadline killed three consecutive job-array windows the same way.Usage
No user-facing change —
ls "$LOG_DIR"before the existence checks in the worker wait loop forces readdir revalidation.Before your PR is "Ready for review"
Pre checks:
Additional Information
Same 32-node job array: startup previously failed 3/3 windows after the first; with this fix all subsequent windows started cleanly and resumed training.