Conversation
g4bl runner integration: - New runner: 'g4bl' field in jobdesc routes through validate_jobdesc to process_g4bl_jobdef in utils/prod_utils.py. Apptainer-wrapped SL7 execution with --cleanenv (env-leak fix: AL9 PYTHONHOME / UPS_DIR / PRODUCTS otherwise leak in and break setupmu2e-art.sh's UPS init). - DEFAULT_G4BL_CONTAINER constant + new poms/g4bl.cfg with SL7 +SingularityImage so grid jobs run directly in SL7 (no nesting). - _is_inside_sl7() runtime detect skips the apptainer wrap when the outer Condor container is already SL7. - JSON shape: runner, embed_dir, main_input, events_per_job, njobs. - Smoke verified: 81KB ROOT written, 0 fatal exceptions. - Files: data/mdc2025/g4bl.json, test/g4bl_smoke_jobdesc.json. Repo hygiene: - Mu2eFilename consolidated to utils/job_common.py (relpathname() method added). utils/datasetFileList.py imports the canonical class. Hash output byte-identical, all 181 unit tests still pass. - Dead code removed from utils/prod_utils.py: parse_jobdef_fields(), make_jobdefs_list() — zero external callers. - Untracked duplicates removed from bin/: listMcsDefs (byte-identical to listRelatedDefs), listRelatedDefs (per option C), and the abandoned smolagents experiment json2jobdef_agent.py + README. - CLAUDE.md prodtools list trimmed accordingly. - .gitignore: root-anchored entries for scratch files (os, sys, test2/, test_runmu2e/, test_reco/, prompts*.txt, mu2e_common.gdml, MDC*-test.json, momentum_resolution_*.png, poms_data.db). Skills: - /parallel-audit: spawn N parallel Explore agents on independent slices, synthesize findings into a prioritized punch list. Inspired by Hermes Agent's subagent-driven-development. Default Mu2e 4-slice cut documented (utils+bin / CLI / DB+JSON / tests+hygiene). Wiki: log entries for skill commissions, repo hygiene, ~/.claude move to /exp/mu2e/app, g4bl runner integration. Raw source staged for the next ingest. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
g4bl runner architecture (was SL7 container, now native AL9): - process_g4bl_jobdef in utils/prod_utils.py simplified ~50 LOC. Was: apptainer exec --cleanenv against fnal-dev-sl7, with HOME / /tmp / embed_dir binds + PYTHONHOME/PATH + --cleanenv to dodge env leaks. Now: bash -c 'unset SPACK_ENV PYTHON*; source mu2e-art.sh; eval $(spack load --sh g4beamline); cd embed_dir; g4bl ...'. No container, no nested wrap — workers run in the standard fnal-wn-el9 outer container set by poms/fermigrid.cfg. - _is_inside_sl7() helper and DEFAULT_G4BL_CONTAINER constant removed. - poms/g4bl.cfg deleted (was the SL7 outer container override; redundant with fermigrid.cfg now). - runmu2e.py refactored: dispatch logic extracted into _dispatch_and_execute() helper (Option E in the architecture discussion); removes the early sys.exit(0) g4bl branch and consolidates the per-runner switch in one function. New build_mu2e_cmd() helper packages the art-side mu2e -c command as a shell=False arg list, closing the shell=True quoting hazard while preserving && semantics. Output naming + map shape: - Output renamed from g4bl.mu2e.* to nts.mu2e.* (canonical Mu2e ntuple tier; matches metacat convention). - desc renamed Mu2EBeamline* -> G4blPOT (descriptive of POT physics, not art-side musing tag). - dsconf -> TESTaa (test marker; clearly out-of-band of MDC campaigns). - POMS map shape minimized to 5 fields (tarball, njobs, runner, inloc, outputs) matching MDC2025-NNN convention. Runtime config (desc, dsconf, main_input, events_per_job) lives in jobpars.json INSIDE the tarball — the tarball is self-describing for grid replay. Embed_dir mode (local smoke) keeps the full schema since there's no tarball. - validate_jobdesc relaxed to require only `outputs` for g4bl tarball mode; full set for embed_dir mode. No fallbacks: missing fields fail loudly per the no-fallbacks rule. Two non-obvious gotchas captured in code comments + memory: - `unset SPACK_ENV` is mandatory before sourcing in subprocess scripts. bin/runmu2e does `muse setup ops` before invoking Python, which activates the ops-019 spack env; subprocess inherits SPACK_ENV; `spack load g4beamline` then searches only ops-019 (no g4beamline) and fails with "matches no installed packages". Mu2e wiki documents this. - g4bl 3.08b on AL9 requires plain `key=value` CLI syntax, NOT `param key=value` — the `param` form is input-file-only. The older 3.08 SL7 build was lenient. Wiki ingest: github.com/fermitools/poms research → wiki/pages/ poms-reference.md (synthesized data model + Mu2e conventions + dispatch lifecycle) + 3 raw sources at wiki/raw/2026-04-28-poms-*. Memories: reference_spack_env_after_muse_setup.md, feedback_no_fallbacks .md, feedback_no_voms_proxy_init.md, reference_g4bl_param_unset_semantics .md. 181/181 unit tests pass. Both smoke modes (embed_dir + tarball) produce SAM-named nts.*.root + log.*.log with 0 fatal exceptions. Demonstrator artifacts staged: tarball at /pnfs/mu2e/tape/phy-etc/cnf/mu2e/G4blPOT/ TESTaa/tar/c7/74/, POMS map at .../poms_map/G4BL-000.json, SAM index def iG4BL-000 (1 file). Pending grid push: cvmfs prodtools deployment + POMS-side stage configuration. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
1 participant
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.
No description provided.