#!/usr/bin/env bash
# Stops one replica by the PID file start-instance.sh wrote for it - never by matching
# the process name or command line, which risks matching the wrong process (including
# this very script's own shell). Usage: stop-instance.sh <port>
#
# Sends SIGTERM, not SIGKILL. "server.shutdown: graceful" in application.yml only
# does anything on SIGTERM: it stops accepting new connections but lets in-flight
# requests finish first. An earlier version used `kill -9` here, which bypasses that
# entirely, and requests that were in flight the instant the process vanished showed
# up in the load generator's summary as ConnectException/IOException - a real
# artifact of skipping the drain step, not a defect in the migration itself. See
# docs/13-graceful-shutdown-vs-kill-9.md.
set -euo pipefail
DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
source "$DIR/env.sh"

PORT="$1"
PID_FILE="$EC_PID_DIR/$PORT.pid"
if [ -f "$PID_FILE" ]; then
    PID="$(cat "$PID_FILE")"
    if kill -0 "$PID" 2>/dev/null; then
        # Deregister BEFORE terminating: tell the load balancer to stop sending new
        # traffic here, then give its health check a couple of poll cycles to notice,
        # THEN stop the process. Skipping this drain window and going straight to
        # SIGTERM is what produced the ConnectException bursts in an earlier run.
        curl -s -X POST "http://localhost:$PORT/admin/drain" -o /dev/null || true
        sleep 1.5
        kill -15 "$PID"
        for i in $(seq 1 40); do
            kill -0 "$PID" 2>/dev/null || break
            sleep 0.25
        done
        if kill -0 "$PID" 2>/dev/null; then
            echo "port $PORT (pid $PID) did not exit gracefully in 10s, sending SIGKILL" >&2
            kill -9 "$PID"
        fi
        echo "stopped port $PORT (pid $PID)"
    fi
    rm -f "$PID_FILE"
fi
wait_down "$PORT" || echo "warning: port $PORT still answering after stop" >&2