padelnomics/deploy.sh

#!/usr/bin/env bash
set -euo pipefail

# ── Ensure sops + age are installed ───────────────────────
APP_DIR="$(cd "$(dirname "$0")" && pwd)"
BIN_DIR="$APP_DIR/bin"
mkdir -p "$BIN_DIR"
export PATH="$BIN_DIR:$PATH"

ARCH=$(uname -m)
case "$ARCH" in
    x86_64)  ARCH_SOPS="amd64"; ARCH_AGE="amd64" ;;
    aarch64) ARCH_SOPS="arm64"; ARCH_AGE="arm64" ;;
    *) echo "Unsupported architecture: $ARCH"; exit 1 ;;
esac

if ! command -v age &>/dev/null; then
    echo "==> Installing age to $BIN_DIR..."
    AGE_VERSION="v1.3.1"
    curl -fsSL "https://dl.filippo.io/age/${AGE_VERSION}?for=linux/${ARCH_AGE}" -o /tmp/age.tar.gz
    tar -xzf /tmp/age.tar.gz -C "$BIN_DIR" --strip-components=1 age/age age/age-keygen
    chmod +x "$BIN_DIR/age" "$BIN_DIR/age-keygen"
    rm /tmp/age.tar.gz
fi

if ! command -v sops &>/dev/null; then
    echo "==> Installing sops to $BIN_DIR..."
    SOPS_VERSION="v3.12.1"
    curl -fsSL "https://github.com/getsops/sops/releases/download/${SOPS_VERSION}/sops-${SOPS_VERSION}.linux.${ARCH_SOPS}" -o "$BIN_DIR/sops"
    chmod +x "$BIN_DIR/sops"
fi

# ── Ensure age keypair exists ─────────────────────────────
AGE_KEY_FILE="${SOPS_AGE_KEY_FILE:-$APP_DIR/age-key.txt}"
export SOPS_AGE_KEY_FILE="$AGE_KEY_FILE"

if [ ! -f "$AGE_KEY_FILE" ]; then
    echo "==> Generating age keypair..."
    age-keygen -o "$AGE_KEY_FILE" 2>&1
    chmod 600 "$AGE_KEY_FILE"
    AGE_PUB=$(grep "public key:" "$AGE_KEY_FILE" | awk '{print $NF}')
    echo ""
    echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
    echo "!! NEW SERVER — add this public key to .sops.yaml:    !!"
    echo "!!                                                     !!"
    echo "!! $AGE_PUB !!"
    echo "!!                                                     !!"
    echo "!! Then run: sops updatekeys .env.prod.sops            !!"
    echo "!! Commit, push, and re-deploy.                        !!"
    echo "!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!"
    echo ""
    exit 1
fi

# ── Decrypt secrets ───────────────────────────────────────
sops --input-type dotenv --output-type dotenv -d .env.prod.sops > .env
chmod 600 .env

COMPOSE="docker compose -f docker-compose.prod.yml"
LIVE_FILE=".live-slot"
ROUTER_CONF="router/default.conf"

# ── Determine slots ─────────────────────────────────────────

CURRENT=$(cat "$LIVE_FILE" 2>/dev/null || echo "none")

if [ "$CURRENT" = "blue" ]; then
    TARGET="green"
else
    TARGET="blue"
fi

echo "==> Current: $CURRENT → Deploying: $TARGET"

# ── Build ───────────────────────────────────────────────────

echo "==> Building $TARGET..."
$COMPOSE --profile "$TARGET" build

# ── Backup DB before migration ────────────────────────────────

BACKUP_TAG="pre-deploy-$(date +%Y%m%d-%H%M%S)"
echo "==> Backing up database (${BACKUP_TAG})..."
$COMPOSE run --rm --entrypoint "" "${TARGET}-app" \
    sh -c "cp /app/data/app.db /app/data/app.db.${BACKUP_TAG} 2>/dev/null || true"

# ── Migrate ─────────────────────────────────────────────────

echo "==> Running migrations..."
$COMPOSE --profile "$TARGET" run --rm "${TARGET}-app" \
    python -m padelnomics.migrations.migrate

# ── Ensure router is healthy before waiting ──────────────────
# nginx -t resolves upstream hostnames — if the config points to a stopped
# slot, the health check fails. Write config for the CURRENT live slot
# (which is still running) so the router stays healthy during --wait.

_write_router_conf() {
    local SLOT="$1"
    mkdir -p "$(dirname "$ROUTER_CONF")"
    cat > "$ROUTER_CONF" <<NGINX
upstream app {
    server ${SLOT}-app:5000;
}

server {
    listen 80;

    location / {
        proxy_pass http://app;
        proxy_set_header Host \$host;
        proxy_set_header X-Real-IP \$remote_addr;
        proxy_set_header X-Forwarded-For \$proxy_add_x_forwarded_for;
        proxy_set_header X-Forwarded-Proto \$http_x_forwarded_proto;
    }
}
NGINX
}

if [ "$CURRENT" != "none" ]; then
    echo "==> Resetting router to current slot ($CURRENT)..."
    _write_router_conf "$CURRENT"
    $COMPOSE restart router
fi

# ── Start & health check ───────────────────────────────────

echo "==> Starting $TARGET (waiting for health check)..."
if ! $COMPOSE --profile "$TARGET" up -d --wait; then
    echo "!!! Health check failed — dumping logs"
    echo "--- ${TARGET}-app logs ---"
    $COMPOSE --profile "$TARGET" logs --tail=60 "${TARGET}-app" 2>&1 || true
    echo "--- router logs ---"
    $COMPOSE logs --tail=10 router 2>&1 || true
    echo "--- litestream logs ---"
    $COMPOSE logs --tail=10 litestream 2>&1 || true
    echo "!!! Rolling back"
    $COMPOSE stop "${TARGET}-app" "${TARGET}-worker" "${TARGET}-scheduler"
    LATEST=$($COMPOSE run --rm --entrypoint "" "${TARGET}-app" \
        sh -c "ls -t /app/data/app.db.pre-deploy-* 2>/dev/null | head -1")
    if [ -n "$LATEST" ]; then
        echo "==> Restoring database from ${LATEST}..."
        $COMPOSE run --rm --entrypoint "" "${TARGET}-app" \
            sh -c "cp '${LATEST}' /app/data/app.db"
    fi
    exit 1
fi

# ── Write router config and reload (new slot is healthy) ────

echo "==> Switching router to $TARGET..."
_write_router_conf "$TARGET"
$COMPOSE exec router nginx -s reload

# ── Cleanup old pre-deploy backups (keep last 3) ─────────────

$COMPOSE run --rm --entrypoint "" "${TARGET}-app" \
    sh -c "ls -t /app/data/app.db.pre-deploy-* 2>/dev/null | tail -n +4 | xargs rm -f" || true

# ── Stop old slot ───────────────────────────────────────────

if [ "$CURRENT" != "none" ]; then
    echo "==> Stopping $CURRENT..."
    $COMPOSE stop "${CURRENT}-app" "${CURRENT}-worker" "${CURRENT}-scheduler"
fi

# ── Record live slot ────────────────────────────────────────

echo "$TARGET" > "$LIVE_FILE"
echo "==> Deployed $TARGET successfully!"