paper-dynasty-discord/health_server.py

"""
HTTP health check server for Paper Dynasty Discord bot.

Provides health and readiness endpoints for container monitoring and orchestration.
"""

import asyncio
import logging
from typing import Optional
from aiohttp import web
from discord.ext import commands

logger = logging.getLogger("discord_app.health")


class HealthServer:
    """HTTP server for health checks and metrics."""

    def __init__(self, bot: commands.Bot, host: str = "0.0.0.0", port: int = 8080):
        """
        Initialize health server.

        Args:
            bot: Discord bot instance to monitor
            host: Host to bind to (default: 0.0.0.0 for container access)
            port: Port to listen on (default: 8080)
        """
        self.bot = bot
        self.host = host
        self.port = port
        self.app = web.Application()
        self.runner: Optional[web.AppRunner] = None
        self.site: Optional[web.TCPSite] = None

        # Setup routes
        self.app.router.add_get("/health", self.health_check)
        self.app.router.add_get("/ready", self.readiness_check)
        self.app.router.add_get("/metrics", self.metrics)
        self.app.router.add_get("/diagnostics", self.diagnostics)

    async def health_check(self, request: web.Request) -> web.Response:
        """
        Basic liveness check - is the process running?

        Returns 200 if the server is responsive.
        """
        return web.json_response(
            {"status": "healthy", "service": "paper-dynasty-discord-bot"}
        )

    async def readiness_check(self, request: web.Request) -> web.Response:
        """
        Readiness check - is the bot ready to serve requests?

        Returns:
            200 if bot is connected to Discord
            503 if bot is not ready
        """
        if self.bot.is_ready():
            return web.json_response(
                {
                    "status": "ready",
                    "discord_connected": True,
                    "latency_ms": (
                        round(self.bot.latency * 1000, 2) if self.bot.latency else None
                    ),
                }
            )
        else:
            return web.json_response(
                {"status": "not_ready", "discord_connected": False}, status=503
            )

    async def metrics(self, request: web.Request) -> web.Response:
        """
        Return bot metrics for monitoring.

        Provides detailed information about bot state for external monitoring systems.
        """
        metrics_data = {
            "bot": {
                "is_ready": self.bot.is_ready(),
                "is_closed": self.bot.is_closed(),
                "latency_ms": (
                    round(self.bot.latency * 1000, 2) if self.bot.latency else None
                ),
            },
            "guilds": {
                "count": len(self.bot.guilds),
                "guild_ids": [g.id for g in self.bot.guilds],
            },
            "users": {"count": len(self.bot.users)},
            "cogs": {"loaded": list(self.bot.cogs.keys()), "count": len(self.bot.cogs)},
        }

        return web.json_response(metrics_data)

    async def diagnostics(self, request: web.Request) -> web.Response:
        """
        Detailed diagnostics for troubleshooting frozen bot.
        Captures state before container restart.
        """
        import sys

        tasks_info = []
        try:
            for task in asyncio.all_tasks():
                tasks_info.append(
                    {
                        "name": task.get_name(),
                        "done": task.done(),
                        "cancelled": task.cancelled(),
                    }
                )
        except Exception as e:
            tasks_info = [f"Error capturing tasks: {e}"]

        diagnostics_data = {
            "bot": {
                "is_ready": self.bot.is_ready(),
                "is_closed": self.bot.is_closed(),
                "latency_ms": (
                    round(self.bot.latency * 1000, 2) if self.bot.latency else None
                ),
            },
            "tasks": {"count": len(tasks_info), "tasks": tasks_info[:20]},
            "cogs": {"loaded": list(self.bot.cogs.keys()), "count": len(self.bot.cogs)},
            "python_version": sys.version,
        }

        return web.json_response(diagnostics_data)

    async def start(self):
        """Start the health check server."""
        self.runner = web.AppRunner(self.app)
        await self.runner.setup()
        self.site = web.TCPSite(self.runner, self.host, self.port)
        await self.site.start()
        logger.info(f"Health check server started on {self.host}:{self.port}")

    async def stop(self):
        """Stop the health check server."""
        if self.site:
            await self.site.stop()
        if self.runner:
            await self.runner.cleanup()
        logger.info("Health check server stopped")


async def run_health_server(bot: commands.Bot, host: str = "0.0.0.0", port: int = 8080):
    """
    Run health server as a background task.

    Args:
        bot: Discord bot instance
        host: Host to bind to
        port: Port to listen on
    """
    server = HealthServer(bot, host, port)
    await server.start()

    # Keep the server running until bot is closed
    try:
        while not bot.is_closed():
            await asyncio.sleep(1)
    finally:
        await server.stop()