hm_ai_qc_report_tool/deploy.sh

#!/bin/bash
# HM AI QC deploy script (Docker Compose).
#
# Usage:
#   deploy.sh dev                 Deploy origin/develop HEAD to this server
#   deploy.sh prod <tag>          Deploy a specific tag to this server
#   deploy.sh dev --dry-run       Show what would change, make no changes
#   deploy.sh prod <tag> --dry-run
#
# Runs on the target server (optical-dev / optical-prod), not your laptop.
# Saves a rollback checkpoint to .last_deploy_rollback before changing
# anything, and auto-rolls back if the post-deploy /health probe fails.
#
# Differences from the AI QC sibling script (intentional):
#   * Docker Compose, not systemd. `docker compose up -d` replaces
#     systemctl restart; `docker compose build` replaces pip install.
#   * `flask db upgrade` runs as a one-shot container before bringing up
#     the web service, so schema changes apply atomically with the deploy.
#   * No "delete frontend / build frontend / copy to /var/www/html" steps
#     from the IT spec — HM QC ships Flask templates, not an SPA bundle.

set -euo pipefail

APP_DIR=${APP_DIR:-/opt/hm-aiqc}
HEALTH_URL=${HEALTH_URL:-http://127.0.0.1:5050/health}
ROLLBACK_FILE="$APP_DIR/.last_deploy_rollback"

MODE=${1:-}
shift || true

DRY_RUN=false
FORCE=false
TARGET_TAG=""

parse_flags() {
    for arg in "$@"; do
        case "$arg" in
            --dry-run) DRY_RUN=true ;;
            --force)   FORCE=true ;;
        esac
    done
}

case "$MODE" in
    dev)
        parse_flags "$@"
        ;;
    prod)
        TARGET_TAG=${1:-}
        shift || true
        parse_flags "$@"
        if [[ -z "$TARGET_TAG" ]]; then
            echo "Usage: $0 prod <tag> [--dry-run] [--force]"
            exit 1
        fi
        ;;
    ""|-h|--help)
        cat <<EOF
Usage:
  $(basename "$0") dev [--dry-run] [--force]
  $(basename "$0") prod <tag> [--dry-run] [--force]

  --dry-run   Show what would change, make no changes.
  --force     Re-build + re-up even if local HEAD already matches the target.
              Implied automatically on first deploy (no running container).

Run on the target server. Requires permission to talk to docker.
EOF
        exit 0
        ;;
    *)
        echo "Unknown mode: $MODE"
        echo "Try: $(basename "$0") --help"
        exit 1
        ;;
esac

cd "$APP_DIR"
if [[ ! -d .git ]]; then
    echo "ERROR: $APP_DIR is not a git repo"
    exit 1
fi

if [[ ! -f .env ]]; then
    echo "ERROR: $APP_DIR/.env not found. Copy from deploy/.env.${MODE}.example and fill in."
    exit 1
fi

CURRENT_REV=$(git rev-parse HEAD)
CURRENT_SHORT=$(git rev-parse --short HEAD)

echo "============================================"
echo "  HM AI QC deploy ($MODE)"
echo "============================================"
echo "Server:  $(hostname)"
echo "Current: $CURRENT_SHORT  $(git log -1 --format='%s' HEAD)"
echo ""
echo "Fetching latest refs..."
git fetch --tags --prune --quiet

if [[ "$MODE" == "dev" ]]; then
    TARGET_REF="origin/develop"
else
    if ! git rev-parse --verify --quiet "refs/tags/$TARGET_TAG^{commit}" > /dev/null; then
        echo "ERROR: Tag '$TARGET_TAG' not found after fetch"
        exit 1
    fi
    TARGET_REF="refs/tags/$TARGET_TAG"
fi

TARGET_REV=$(git rev-parse "$TARGET_REF")
TARGET_SHORT=$(git rev-parse --short "$TARGET_REF")

CONTAINER_RUNNING=false
if docker compose ps --status running --quiet web 2>/dev/null | grep -q .; then
    CONTAINER_RUNNING=true
fi

if [[ "$CURRENT_REV" == "$TARGET_REV" ]]; then
    if [[ "$CONTAINER_RUNNING" == "false" ]]; then
        echo "Already at $TARGET_SHORT, but no running container — proceeding to build/up."
        FORCE=true
    elif [[ "$FORCE" == "true" ]]; then
        echo "Already at $TARGET_SHORT — --force passed, re-building and restarting."
    else
        echo "Already at $TARGET_SHORT — nothing to do. (Pass --force to rebuild anyway.)"
        exit 0
    fi
fi

echo "Target:  $TARGET_SHORT  $(git log -1 --format='%s' "$TARGET_REF")"
echo ""

if [[ "$CURRENT_REV" != "$TARGET_REV" ]]; then
    echo "Commits to apply:"
    git log --oneline "$CURRENT_REV..$TARGET_REV" | head -20
    CHANGE_COUNT=$(git log --oneline "$CURRENT_REV..$TARGET_REV" | wc -l | tr -d ' ')
    if [[ $CHANGE_COUNT -gt 20 ]]; then
        echo "  ... and $((CHANGE_COUNT - 20)) more"
    fi
    echo ""

    if git diff --name-only "$CURRENT_REV" "$TARGET_REV" | grep -qE "(^|/)migrations/versions/"; then
        echo "Note: Alembic migrations changed — flask db upgrade will run."
        echo ""
    fi
fi

if [[ "$DRY_RUN" == "true" ]]; then
    echo "Dry run — no changes made."
    exit 0
fi

read -r -p "Proceed with deploy? (y/N): " confirm
if [[ ! $confirm =~ ^[Yy]$ ]]; then
    echo "Cancelled."
    exit 0
fi

echo "$CURRENT_REV" > "$ROLLBACK_FILE"

echo "Applying changes..."
git reset --hard "$TARGET_REV"

echo "Building images..."
docker compose build

echo "Starting services (entrypoint runs flask db upgrade first)..."
docker compose up -d

# Poll $HEALTH_URL every 2s until it answers 2xx, or timeout.
# 60s window allows for migration time on first boot of a major release.
wait_for_health() {
    local max_attempts=30   # 30 × 2s = 60s window
    for ((i=1; i<=max_attempts; i++)); do
        sleep 2
        if curl -sf -o /dev/null "$HEALTH_URL"; then
            echo "  healthy after ${i}x2s"
            return 0
        fi
    done
    return 1
}

echo "Smoke testing $HEALTH_URL..."
if wait_for_health; then
    NEW_SHORT=$(git rev-parse --short HEAD)
    echo ""
    echo "Deploy OK. Now at $NEW_SHORT."
    echo "Rollback target saved: $CURRENT_SHORT (run deploy/rollback.sh last to revert)"
    exit 0
fi

echo ""
echo "Smoke test failed after 60s — rolling back to $CURRENT_SHORT..."
git reset --hard "$CURRENT_REV"
docker compose build
docker compose up -d

if wait_for_health; then
    echo "Rolled back successfully. Service healthy at $CURRENT_SHORT."
    echo "Investigate: docker compose logs --tail=200 web"
    exit 1
fi

echo "ROLLBACK ALSO FAILED. Service is in a broken state."
echo "docker compose ps"
echo "docker compose logs --tail=200 web"
exit 2