Repository navigation
CI: run the light IOTests jobs on the self-hosted alma10 runners - #322
Merged
Merged
Conversation
…alma10 runners On a self-hosted runner restore_heptools no longer uses the GitHub caches (built on ubuntu, with /home/runner baked into the install): each machine builds the missing HEPTools components once into its own ~/.cache/HEPtools (selfhosted_heptools). The restore_heptools_* actions use $HOME instead of /home/runner and find the lhapdf/rivet/contur python modules on other images. Fork PRs and the two matrix jobs stay on ubuntu-24.04.
Their makefiles do not declare the Fortran module dependencies, so the -j$(nproc) inherited from MAKEFLAGS failed on avh_olo_foriregi_dp_box.mod.
-j$(nproc) got the runner service OOM-killed on the alma10 machines while building rivet; use one compile per 2 GB of MemAvailable, at most one per core.
The alma10 machines are small (2 cores, 3.3 GB). Keep on ubuntu-24.04 the jobs that need rivet/contur or the NLO/loop libraries (IOtest_bundle_2/3, acceptancetest_0, madevent bundles 1/2/3/6); the moved madevent jobs ask restore_heptools (new selfhosted_tools input) to build only lhapdf, boost, pythia8 and eMELA on the machine.
…ma10 Only IOtest_bundle_1, IOtest_loop_sqso and IOtest_gnuplot run on the alma10 runners, and none of them needs HEPTools, so the per-machine HEPTools build (selfhosted_heptools) and the restore_heptools changes are dropped again.
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Moves three IOTests jobs to the self-hosted alma10 runners (cern1–cern5), with the same runner choice as the merged #321: fork PRs stay on
ubuntu-24.04.IOtest_bundle_1IOtest_loop_sqsorestore_ufomodel)IOtest_gnuplotrestore-pip-cache)IOtest_bundle_2/IOtest_bundle_3(NLO/loop libraries viarestore_all) and the whole madevent acceptance suite stay on ubuntu.Why only these. The alma10 machines are small (2 cores, 3.3 GB RAM, 8 GB swap): jobs run ~1.5x slower than on a hosted runner, and a
make -jrivet build OOM-killed a runner service. Porting the jobs that need HEPTools was tried on this branch (a per-machine HEPTools build into~/.cache/HEPtools, since the GitHub caches carry/home/runnerpaths) and dropped again: those jobs would have needed rivet/contur and the loop libraries built on every machine, and their load (~16% of the hosted minutes per push) would saturate the five machines on busy days. The commits stay in this branch's history for when the machines are bigger; the net diff isIOtest.ymlonly.Load. Together with #321, about 64 of ~700 hosted runner-minutes per push (~9%) move to alma10.