Let learned resource estimates replace dynamic sampling in anchorMC - #2439
Open
sawenzel wants to merge 1 commit into
Open
Let learned resource estimates replace dynamic sampling in anchorMC#2439sawenzel wants to merge 1 commit into
sawenzel wants to merge 1 commit into
Conversation
This adds an ALIEN_O2DPG_RESOURCES hook to anchorMC.sh so a production can run with the resource estimates a pilot job measured. - anchorMC.sh passed --dynamic-resources unconditionally, so learned estimates could only be added on top of runtime sampling, never instead of it. - When ALIEN_O2DPG_RESOURCES names a file, --update-resources is passed and --dynamic-resources is not; without it the invocation is unchanged. - The file is the json-stat output of MC/utils/o2dpg_sim_metrics.py. - ALIEN_O2DPG_FILEGRAPH is added to the usage text, which did not mention it. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
REQUEST FOR PRODUCTION RELEASES: This will add The following labels are available |
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
anchorMC.shpasses--dynamic-resourcesunconditionally, so resource estimates learned by a pilot job could only ever be added on top of runtime sampling, never used instead of it. This adds anALIEN_O2DPG_RESOURCEShook, mirroring the existingALIEN_O2DPG_FILEGRAPHone.When the variable names a file,
--update-resources <file>is passed and--dynamic-resourcesis not. Without it the invocation is byte-for-byte what it was.The file is the
json-statoutput ofMC/utils/o2dpg_sim_metrics.py:Tested on the GRID
Nine anchored jobs (2024 pp
apass1, run 553185, 2 TF) pinned toALICE::CERN::CERN-SIRIUS, in three arms of three: legacy runner with dynamic sampling, new runner with dynamic sampling, and new runner with learned estimates plus early file removal. All nine returnedANCHORMC_RC=0.With the hook set, the runner logs the injection actually taking effect — 74 MEM, 74 CPU and 46 walltime overrides, replacing the workflow's placeholder defaults:
Walltime could not distinguish the arms: the within-arm spread (617–1887 s in one arm at fixed
SPLITIDon one CE) is far larger than any difference between arm means (1041 / 1106 / 916 s). That is a statement about GRID variance, not about this change.🤖 Generated with Claude Code