Skip to content

Stop re-downloading cached FWL_DATA and AGNI data on every CI run #794

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
nichollsh merged 3 commits into from
Jul 29, 2026
Merged

Stop re-downloading cached FWL_DATA and AGNI data on every CI run #794

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0e06eb5
Key the FWL_DATA cache on the fwl-io manifest so re-pins refresh it
egpbos Jul 28, 2026
d7a4ed9
Skip the AGNI data fetch when the cached res/ tree is complete
egpbos Jul 28, 2026
70c1ba7
Merge branch 'main' into speed-up-fast-pr-checks-caching
nichollsh Jul 29, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
23 changes: 22 additions & 1 deletion .github/actions/setup-proteus/action.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -387,6 +387,18 @@ runs:
fi
return $rc
}
# Skip the fetch when a warm cache already holds a complete res/ tree.
# res_complete asserts every required artifact is present AND non-empty
# (a zero-byte file from a green-but-empty fetch fails the -s test), so a
# tree that passes needs no repair and re-pulling it is pure download
# cost (~200 s, dominated by the thermodynamics archive). An incomplete
# or partially-written tree fails the check and falls through to the
# retry loop below, which overwrites and repairs it, so the repair path
# the unconditional fetch was added for is preserved.
if res_complete 2>/dev/null; then
echo "AGNI basic data already complete in res/ (warm cache); skipping fetch."
exit 0
fi
fetch="$GITHUB_WORKSPACE/AGNI/src/get_data.sh"
delays=(10 30 60)
attempts=$(( ${#delays[@]} + 1 ))
Expand Down Expand Up @@ -527,11 +539,20 @@ runs:
# ------------------------------------------------------------------
# 9. FWL_DATA cache + smoke / full data download
# ------------------------------------------------------------------
# The key hashes BOTH dataset descriptors: .github/data-manifest.yaml (the
# older download() datasets: spectral files, stellar spectra, tracks) and
# src/proteus/data/proteus_manifest.toml (the fwl-io pins for the exoplanet
# + mass-radius reference data). Keying on the fwl-io manifest as well means
# a Zenodo re-pin busts the cache and forces a fresh save. Without it, files
# pinned after the cache was first frozen re-download on every run: the key
# stays constant, actions/cache exact-hits and never re-saves, so the stale
# tree can never pick them up. The restore-keys prefix still lets a fresh
# save build on the previous tree instead of re-downloading everything.
- name: Cache FWL_DATA
uses: actions/cache@v5
with:
path: ~/fwl_data
key: fwl-data-${{ runner.os }}-${{ hashFiles('.github/data-manifest.yaml') }}
key: fwl-data-${{ runner.os }}-${{ hashFiles('.github/data-manifest.yaml', 'src/proteus/data/proteus_manifest.toml') }}
restore-keys: |
fwl-data-${{ runner.os }}-

Expand Down
7 changes: 4 additions & 3 deletions .github/data-manifest.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,9 +1,10 @@
# PROTEUS data manifest -- explicit version pins for the datasets CI caches.
#
# The CI composite action (.github/actions/setup-proteus/action.yml) hashes
# this file to key the FWL_DATA cache. The cache invalidates only when this
# manifest changes; CI runs that don't change the manifest reuse the
# previously cached FWL_DATA tree.
# this file together with src/proteus/data/proteus_manifest.toml (the fwl-io
# dataset pins) to key the FWL_DATA cache. The cache invalidates when either
# descriptor changes; CI runs that change neither reuse the previously cached
# FWL_DATA tree.
#
# Bump a version string in this file when:
# * a new dataset becomes a hard requirement of a tier the CI runs;
Expand Down
Loading