Deploying voyager 3 without its edge worker: the no-edge marker, history replayed into the catalog
Test and deploy / test (push) Failing after 51s
Test and deploy / edge (push) Successful in 1m52s
Test and deploy / deploy (push) Skipped

deploy.sh and the nightly resume leave the edge stopped while .cache/no-edge
exists (range by range it would split releases at every edge with no merger
left). deploy/replay-window.sh takes DB and OUT and passes
REPLAY_INTO_CATALOG=1, which lets replay:corpus write a database not named
*_replay or *_scratch (never the test one).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
This commit is contained in:
Daan MeijerandClaude Fable 5.1 committed 2026-10-26 16:30:56 +01:00
1 parent 91984d3625
commit 8f5f83b4ea
5 files changed
+34 -13

No files matched your search

+8 -1
View File
@@ -51,7 +51,14 @@ if [ -e "$CACHE/paused" ]; then
docker compose up -d --no-deps node-api web
else
docker compose up -d --no-deps node-api web coordinator prober merger judge spotter spot-inspector
docker compose up -d --no-deps edge
if [ -e "$CACHE/no-edge" ]; then
# Voyager 3 (2026-10-26, LOG.md (116)): the worker indexes range by range, without the carried window, and
# would split releases at every range edge with nothing to join them; history comes from the corpus through
# deploy/replay-window.sh until the window worker is built. Remove the marker to start it again.
echo "== no-edge: the edge worker left stopped"
else
docker compose up -d --no-deps edge
fi
fi
echo "$SHA $(date -Is)" >> "$ROOT/deployed.log"
# The images this replaced (321 MB each; the root filesystem is the NVMe, with a 10 GB floor).