ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

docker: detach the notebook network refresh from container startup

Browse source 
The GitHub refresh phase ran synchronously in the entrypoint's notebook
sync, so an offline or slow network could hold container startup for up
to two fetch timeouts (ls-remote + clone, about two minutes at the
defaults) despite the sync being described as non-blocking. The local
template populate and the categorized view still run in the foreground;
the refresh now re-enters itself as a detached child (guarded by a flag
so it forks once), whose phase-1 pass no-ops via the hash state and
whose finalize is idempotent. Verified with an unreachable remote and
an 8 second timeout: the parent returns in under a second with the
notebooks populated while the child owns the waiting.
This commit is contained in:
Daniel Han 2026-07-20 06:04:49 +00:00
commit f19ef1cfd2
1 changed files with 8 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

8
docker/unsloth_sync_notebooks.sh
View file

@ -172,9 +172,17 @@ if [ -f "$STATE" ] && [ "${UNSLOTH_KEEP_DELETED_NOTEBOOKS:-0}" != "1" ]; then
fi
# 2) Best-effort GitHub refresh -- only when upstream has advanced. Edits win.
# Detached: the local populate above already ran, and the refresh can spend up
# to 2x TIMEOUT on ls-remote + clone when offline, which must not delay
# container startup. The child re-enters past phase 1 (hash state makes it a
# no-op) and the flag keeps it from forking again.
[ "${UNSLOTH_SKIP_NOTEBOOK_REFRESH:-0}" = "1" ] && exit 0
command -v git >/dev/null 2>&1 || exit 0
command -v sha256sum >/dev/null 2>&1 || exit 0
if [ "${UNSLOTH_NB_REFRESH_CHILD:-0}" != "1" ]; then
UNSLOTH_NB_REFRESH_CHILD=1 "$0" >/dev/null 2>&1 &
exit 0
fi
last="$(cat "$SYNCED" 2>/dev/null || true)"
remote="$(timeout "$TIMEOUT" git ls-remote "$REMOTE" HEAD 2>/dev/null | cut -f1)"