Files
hechakuaa_bot/health.py

260 lines
11 KiB
Python

"""
health.py — Servidor HTTP para health check y webhook de Telegram.
Rutas:
GET /health → estado mínimo (público) | estado completo (con Bearer token)
GET /health/history → últimos N eventos del audit_log — REQUIERE Bearer token
POST /webhook → recibe updates de Telegram (solo en modo webhook)
Seguridad:
- /health sin auth: solo devuelve {"status": "ok"|"degraded", "timestamp": "..."}
Sin conteos de tenants ni timing que revele inteligencia operacional.
- /health con auth: respuesta completa con métricas internas.
- /health/history siempre requiere auth.
- chat_ids en el historial se enmascaran como hashes opacos (no reversibles desde HTTP).
- Webhook valida X-Telegram-Bot-Api-Secret-Token con compare_digest.
Configurar en .env:
HEALTH_AUTH_TOKEN=<token largo aleatorio>
(generar con: python -c "import secrets; print(secrets.token_hex(32))")
"""
from __future__ import annotations
import asyncio
import hashlib
import json
import secrets
from datetime import datetime, timedelta, timezone
from zoneinfo import ZoneInfo
_PY_TZ = ZoneInfo("America/Asuncion")
from typing import TYPE_CHECKING
import aiohttp.web
import structlog
from telegram import Update
if TYPE_CHECKING:
from config import Settings
from telegram.ext import Application
logger = structlog.get_logger(__name__)
# ── Estado global del poller ──────────────────────────────────────────────────
_last_poll_at: datetime | None = None
_last_poll_tenant_count: int = 0
_alert_sent: bool = False
def record_poll(tenant_count: int) -> None:
"""Llamar desde poller.py después de cada ronda automática."""
global _last_poll_at, _last_poll_tenant_count, _alert_sent
_last_poll_at = datetime.now(timezone.utc)
_last_poll_tenant_count = tenant_count
_alert_sent = False
def _mask_chat_id(chat_id: int | None) -> str | None:
"""
Convierte un chat_id en un identificador opaco no reversible.
Preserva la capacidad de correlacionar eventos del mismo usuario
sin exponer el Telegram ID real.
"""
if chat_id is None:
return None
return "usr_" + hashlib.sha256(str(chat_id).encode()).hexdigest()[:8]
# ── Servidor ──────────────────────────────────────────────────────────────────
class HealthServer:
def __init__(self, settings: Settings, application: Application) -> None:
self._settings = settings
self._app_ptb = application
self._runner: aiohttp.web.AppRunner | None = None
self._watcher_task: asyncio.Task | None = None
async def start(self) -> None:
app = aiohttp.web.Application()
app.router.add_get("/health", self._handle_health)
app.router.add_get("/health/history", self._handle_history)
if self._settings.webhook_url:
app.router.add_post("/webhook", self._handle_webhook)
logger.info("webhook_endpoint_activo", url=f"{self._settings.webhook_url}/webhook")
self._runner = aiohttp.web.AppRunner(app)
await self._runner.setup()
site = aiohttp.web.TCPSite(self._runner, "0.0.0.0", self._settings.health_port)
await site.start()
logger.info(
"health_server_iniciado",
port=self._settings.health_port,
auth_configurada=bool(self._settings.health_auth_token),
)
self._watcher_task = asyncio.create_task(self._watcher_loop())
async def stop(self) -> None:
if self._watcher_task:
self._watcher_task.cancel()
try:
await self._watcher_task
except asyncio.CancelledError:
pass
if self._runner:
await self._runner.cleanup()
logger.info("health_server_detenido")
# ── Auth ──────────────────────────────────────────────────────────────────
def _is_authenticated(self, request: aiohttp.web.Request) -> bool:
"""
Valida el header Authorization: Bearer <token>.
Si health_auth_token no está configurado, siempre retorna False
(los endpoints protegidos devuelven respuesta mínima, no error).
"""
if not self._settings.health_auth_token:
return False
auth = request.headers.get("Authorization", "")
if not auth.startswith("Bearer "):
return False
provided = auth[7:]
# compare_digest previene timing attacks sobre el token de admin
return secrets.compare_digest(provided, self._settings.health_auth_token)
# ── Handlers ─────────────────────────────────────────────────────────────
async def _handle_health(self, request: aiohttp.web.Request) -> aiohttp.web.Response:
now = datetime.now(timezone.utc)
stale = False
if _last_poll_at:
stale = (now - _last_poll_at) > timedelta(hours=self._settings.health_alert_hours)
# Respuesta pública mínima — no revela conteos ni timing operacional
data: dict = {
"status": "degraded" if stale else "ok",
"timestamp": now.isoformat(),
}
# Respuesta extendida solo con auth válida
if self._is_authenticated(request):
from database import get_conn
conn = get_conn()
async with conn.execute(
"SELECT COUNT(*) as n FROM tenants WHERE estado = 'activo'"
) as cur:
row = await cur.fetchone()
tenant_count = row["n"] if row else 0
seconds_since_poll: int | None = None
if _last_poll_at:
seconds_since_poll = int((now - _last_poll_at).total_seconds())
data.update({
"tenants_activos": tenant_count,
"last_poll": _last_poll_at.isoformat() if _last_poll_at else None,
"last_poll_tenants": _last_poll_tenant_count,
"seconds_since_last_poll": seconds_since_poll,
"poll_interval_minutes": self._settings.poll_interval_minutes,
"webhook_mode": bool(self._settings.webhook_url),
})
return aiohttp.web.Response(
text=json.dumps(data, indent=2),
content_type="application/json",
status=503 if stale else 200,
)
async def _handle_history(self, request: aiohttp.web.Request) -> aiohttp.web.Response:
# Este endpoint siempre requiere autenticación — contiene PII (chat_ids)
if not self._is_authenticated(request):
logger.warning(
"health_history_acceso_denegado",
ip=request.remote,
ua=request.headers.get("User-Agent", "")[:80],
)
return aiohttp.web.Response(
status=401,
text=json.dumps({"error": "Authentication required"}),
content_type="application/json",
headers={"WWW-Authenticate": 'Bearer realm="pedrito-health"'},
)
from database import get_conn
conn = get_conn()
try:
limit = min(int(request.rel_url.query.get("limit", "50")), 200)
except (ValueError, TypeError):
limit = 50
async with conn.execute(
"SELECT timestamp, timestamp_local, chat_id, evento, resultado, duracion_ms "
"FROM audit_log ORDER BY id DESC LIMIT ?",
(limit,),
) as cur:
rows = await cur.fetchall()
# Enmascarar chat_ids: el consumidor puede correlacionar eventos del mismo usuario
# pero no puede determinar la identidad real sin acceso a la DB.
records = []
for r in rows:
row_dict = dict(r)
row_dict["chat_id"] = _mask_chat_id(row_dict.get("chat_id"))
records.append(row_dict)
return aiohttp.web.Response(
text=json.dumps(records, indent=2),
content_type="application/json",
)
async def _handle_webhook(self, request: aiohttp.web.Request) -> aiohttp.web.Response:
provided = request.headers.get("X-Telegram-Bot-Api-Secret-Token", "")
if self._settings.webhook_secret:
if not secrets.compare_digest(provided, self._settings.webhook_secret):
logger.warning("webhook_token_invalido", ip=request.remote)
# Respuesta genérica — no revelar si es auth vs otro error
return aiohttp.web.Response(status=404, text="Not Found")
try:
body = await request.json()
except Exception:
return aiohttp.web.Response(status=400, text="Bad Request")
update = Update.de_json(data=body, bot=self._app_ptb.bot)
await self._app_ptb.update_queue.put(update)
return aiohttp.web.Response(status=200, text="OK")
# ── Watcher ───────────────────────────────────────────────────────────────
async def _watcher_loop(self) -> None:
"""Alerta al admin si el poller lleva más de health_alert_hours sin actualizar."""
global _alert_sent
while True:
await asyncio.sleep(300)
if _last_poll_at is None or _alert_sent:
continue
ahora_py = datetime.now(_PY_TZ)
en_horario = (
ahora_py.weekday() < 5
and self._settings.poll_hora_inicio <= ahora_py.hour < self._settings.poll_hora_fin
)
if not en_horario:
continue
delta = datetime.now(timezone.utc) - _last_poll_at
if delta <= timedelta(hours=self._settings.health_alert_hours):
continue
minutos = int(delta.total_seconds() // 60)
try:
await self._app_ptb.bot.send_message(
chat_id=self._settings.telegram_chat_id,
text=(
f"⚠️ El poller no actualizó hace {minutos} minutos.\n"
"Puede haber un problema con el bot o la conexión al PJ."
),
)
_alert_sent = True
logger.warning("health_alerta_enviada", minutos=minutos)
except Exception as e:
logger.error("health_alerta_error", error=str(e))