[release/cvs-0.2.0] cvs/rccl: fix dmesg false positives and per-user RCCL hostfile#170
Open
speriaswamy-amd wants to merge 1 commit into
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
Cherry-pick of #169 onto
release/cvs-0.2.0. Three small fixes that came out of debugging a 28-noderccl_perfrun on a shared cluster:cvs/lib/verify_lib.py— dropRunlist is getting oversubscribedandExpect reduced ROCm performancefrom thedriverfailure pattern. These are amdgpu kernel info-level messages that fire routinely on large multi-rank RCCL launches even when the run is healthy. Real driver-side errors (Queue preemption failed,Failed to evict process queues,No more SDMA queue to allocate,amdgpu: process pid) are still matched.cvs/tests/rccl/rccl_perf.py— flipverify_dmesg_for_errors(..., till_end_flag=False)so the dmesg scan is bounded by each parametrized test's ownstart_time/end_timewindow. Withtill_end_flag=Truea kernel event from one test (e.g. ascatter_perfsegfault) was failing every subsequent parametrized test for the rest of the run.cvs/lib/rccl_lib.py— write the mpirun hostfile to/tmp/rccl_hosts_file_<USER>.txtinstead of the shared/tmp/rccl_hosts_file.txt, so two users on the same shared cluster nodes don't collide. Replaces thesudo rm -fworkaround on this branch.Conflict resolution notes
cvs/lib/rccl_lib.py: this branch had asudo rm -f /tmp/rccl_hosts_file.txtworkaround for the same multi-user collision. The per-user path supersedes it; workaround removed.cvs/tests/rccl/rccl_perf.py:maingatesverify_dmesg_for_errorsonif can_use_sudo:(passwordless-sudo guard not yet on this branch). For this cherry-pick the call stays unconditional — only thetill_end_flagvalue is changed — to avoid introducing an undefinedcan_use_sudoreference.cvs/lib/verify_lib.py: auto-merged cleanly.Test plan
main(18+/8- across 3 files).python -c "import ast; ast.parse(...)"on all three files passes.rccl_perffrom this branch on the validation cluster.Made with Cursor