worldquant-alpha-system/scripts/deploy-production.sh

#!/usr/bin/env bash
# Deploy on the target Linux Docker host with configuration injected by Gitea.
# Returns nonzero on configuration/build/migration/health failure. Never removes data.
set -Eeuo pipefail
umask 077

cd "$(dirname "${BASH_SOURCE[0]}")/.."
# Fail before touching the host; never read a checkout's local .env file.
for key in WQ_EMAIL WQ_PASSWORD DATABASE_URL ADMIN_PASSWORD ENCRYPTION_KEY PUBLIC_ORIGIN; do
  if [[ -z "${!key:-}" ]]; then
    echo "Missing required Gitea configuration: $key" >&2
    exit 1
  fi
done
export DEPLOY_TAG="${DEPLOY_TAG:-$(git rev-parse HEAD)}"
compose=(docker compose --env-file /dev/null -p wq-alpha-production -f compose.production.yaml)
lock_id=""
cleanup() {
  local rc=$?
  "${compose[@]}" --profile jobs ps -a || true
  # Remove only the lock acquired by this process, never another deployment's lock.
  if [[ -n "$lock_id" ]]; then
    docker rm "$lock_id" >/dev/null || true
  fi
  exit "$rc"
}
trap cleanup EXIT
trap 'exit 130' INT
trap 'exit 143' TERM

"${compose[@]}" config --quiet
"${compose[@]}" --profile jobs config --quiet
"${compose[@]}" build backend web
# Docker enforces unique container names across runner jobs and checkout paths.
# The lock container is never started and receives no deployment credentials.
if ! lock_id=$(docker create --name wq-alpha-production-deploy-lock \
  --label "wq.deploy.commit=$DEPLOY_TAG" \
  --network none --entrypoint /bin/true "wq-alpha-production-backend:$DEPLOY_TAG"); then
  echo 'Cannot acquire deployment lock; check Docker and other active deployments.' >&2
  exit 1
fi
# Validate secrets and DB connectivity before interrupting the running version.
"${compose[@]}" run --rm --no-deps backend python -c '
import asyncio
from app.config import Settings
from sqlalchemy import text
from sqlalchemy.ext.asyncio import create_async_engine
async def check():
    settings = Settings()
    engine = create_async_engine(settings.database_url)
    try:
        async with engine.connect() as connection:
            await connection.execute(text("SELECT 1"))
    finally:
        await engine.dispose()
try:
    asyncio.run(check())
except Exception:
    raise SystemExit("Production configuration/database preflight failed; check env and database access.") from None
'

# Record immutable image references before switching; do not prune old images.
previous_images=$("${compose[@]}" images --quiet)
if [[ -n "$previous_images" ]]; then
  echo "Previous deployment images (retain for rollback):"
  docker image inspect --format '{{.Id}} {{json .RepoTags}}' $previous_images
fi
"${compose[@]}" stop web backend
"${compose[@]}" --profile jobs run --rm --no-deps migrate
"${compose[@]}" up -d --no-build --remove-orphans --wait --wait-timeout 180 backend web
echo "Production is healthy; release $DEPLOY_TAG"