anderson-ufrj
refactor(performance): replace all json imports with json_utils
9730fbc
"""
Comprehensive health check system with dependency monitoring.
This module provides advanced health checking capabilities for all
system dependencies and components.
"""
import asyncio
import time
from typing import Dict, Any, List, Optional, Callable, Union
from datetime import datetime, timedelta
from enum import Enum
from dataclasses import dataclass, field
from src.core import json_utils
import httpx
import redis.asyncio as redis
from sqlalchemy import text
from sqlalchemy.ext.asyncio import AsyncSession
from src.core import get_logger, settings
from src.infrastructure.observability import BusinessMetrics, get_structured_logger
logger = get_structured_logger(__name__, component="health_checker")
class HealthStatus(str, Enum):
"""Health status levels."""
HEALTHY = "healthy"
DEGRADED = "degraded"
UNHEALTHY = "unhealthy"
UNKNOWN = "unknown"
class DependencyType(str, Enum):
"""Types of dependencies."""
DATABASE = "database"
CACHE = "cache"
EXTERNAL_API = "external_api"
MESSAGE_QUEUE = "message_queue"
FILE_SYSTEM = "file_system"
LLM_SERVICE = "llm_service"
INTERNAL_SERVICE = "internal_service"
@dataclass
class HealthCheckResult:
"""Result of a health check operation."""
name: str
status: HealthStatus
dependency_type: DependencyType
response_time_ms: float
message: str
details: Dict[str, Any] = field(default_factory=dict)
timestamp: datetime = field(default_factory=datetime.utcnow)
error: Optional[str] = None
def to_dict(self) -> Dict[str, Any]:
"""Convert to dictionary."""
return {
"name": self.name,
"status": self.status.value,
"dependency_type": self.dependency_type.value,
"response_time_ms": self.response_time_ms,
"message": self.message,
"details": self.details,
"timestamp": self.timestamp.isoformat(),
"error": self.error
}
class HealthCheckConfig:
"""Configuration for health checks."""
def __init__(
self,
timeout_seconds: float = 5.0,
warning_threshold_ms: float = 1000.0,
critical_threshold_ms: float = 3000.0,
max_retries: int = 2,
retry_delay_seconds: float = 0.5
):
self.timeout_seconds = timeout_seconds
self.warning_threshold_ms = warning_threshold_ms
self.critical_threshold_ms = critical_threshold_ms
self.max_retries = max_retries
self.retry_delay_seconds = retry_delay_seconds
class BaseHealthCheck:
"""Base class for health checks."""
def __init__(
self,
name: str,
dependency_type: DependencyType,
config: Optional[HealthCheckConfig] = None
):
self.name = name
self.dependency_type = dependency_type
self.config = config or HealthCheckConfig()
self.logger = get_structured_logger(f"health.{name}")
async def check(self) -> HealthCheckResult:
"""Perform health check with retries."""
last_error = None
for attempt in range(self.config.max_retries + 1):
try:
start_time = time.time()
# Perform the actual health check
result = await asyncio.wait_for(
self._perform_check(),
timeout=self.config.timeout_seconds
)
response_time_ms = (time.time() - start_time) * 1000
# Determine status based on response time
if response_time_ms > self.config.critical_threshold_ms:
status = HealthStatus.UNHEALTHY
message = f"Response time {response_time_ms:.2f}ms exceeds critical threshold"
elif response_time_ms > self.config.warning_threshold_ms:
status = HealthStatus.DEGRADED
message = f"Response time {response_time_ms:.2f}ms exceeds warning threshold"
else:
status = HealthStatus.HEALTHY
message = "Dependency is healthy"
# Override status if check returned specific status
if hasattr(result, 'status'):
status = result.status
message = result.get('message', message)
health_result = HealthCheckResult(
name=self.name,
status=status,
dependency_type=self.dependency_type,
response_time_ms=response_time_ms,
message=message,
details=result if isinstance(result, dict) else {}
)
# Log the result
self.logger.info(
f"Health check passed: {self.name}",
operation="health_check",
duration_ms=response_time_ms,
status=status.value,
attempt=attempt + 1
)
return health_result
except asyncio.TimeoutError:
last_error = f"Health check timed out after {self.config.timeout_seconds}s"
self.logger.warning(
f"Health check timeout: {self.name}",
operation="health_check",
timeout_seconds=self.config.timeout_seconds,
attempt=attempt + 1
)
except Exception as e:
last_error = str(e)
self.logger.error(
f"Health check failed: {self.name}",
operation="health_check",
error=e,
attempt=attempt + 1
)
# Wait before retry
if attempt < self.config.max_retries:
await asyncio.sleep(self.config.retry_delay_seconds)
# All attempts failed
return HealthCheckResult(
name=self.name,
status=HealthStatus.UNHEALTHY,
dependency_type=self.dependency_type,
response_time_ms=self.config.timeout_seconds * 1000,
message=f"Health check failed after {self.config.max_retries + 1} attempts",
error=last_error
)
async def _perform_check(self) -> Union[Dict[str, Any], bool]:
"""Override this method to implement specific health check logic."""
raise NotImplementedError("Subclasses must implement _perform_check")
class DatabaseHealthCheck(BaseHealthCheck):
"""Health check for database connectivity."""
def __init__(self, session_factory: Callable, config: Optional[HealthCheckConfig] = None):
super().__init__("database", DependencyType.DATABASE, config)
self.session_factory = session_factory
async def _perform_check(self) -> Dict[str, Any]:
"""Check database connectivity and basic operations."""
async with self.session_factory() as session:
# Test basic connectivity
result = await session.execute(text("SELECT 1"))
row = result.fetchone()
if row[0] != 1:
raise Exception("Database query returned unexpected result")
# Get database stats
stats_result = await session.execute(text("""
SELECT
count(*) as active_connections,
(SELECT setting FROM pg_settings WHERE name = 'max_connections') as max_connections
FROM pg_stat_activity
WHERE state = 'active'
"""))
stats = stats_result.fetchone()
return {
"connection_status": "active",
"active_connections": stats[0],
"max_connections": int(stats[1]),
"utilization": stats[0] / int(stats[1]) if stats[1] else 0
}
class RedisHealthCheck(BaseHealthCheck):
"""Health check for Redis connectivity."""
def __init__(self, redis_url: str, config: Optional[HealthCheckConfig] = None):
super().__init__("redis", DependencyType.CACHE, config)
self.redis_url = redis_url
async def _perform_check(self) -> Dict[str, Any]:
"""Check Redis connectivity and basic operations."""
redis_client = redis.from_url(self.redis_url)
try:
# Test basic operations
test_key = "health_check_test"
test_value = f"test_{int(time.time())}"
# Set and get test value
await redis_client.set(test_key, test_value, ex=60)
retrieved_value = await redis_client.get(test_key)
if retrieved_value.decode() != test_value:
raise Exception("Redis get/set test failed")
# Clean up test key
await redis_client.delete(test_key)
# Get Redis info
info = await redis_client.info()
return {
"connection_status": "active",
"version": info.get("redis_version"),
"used_memory": info.get("used_memory_human"),
"connected_clients": info.get("connected_clients"),
"operations_per_sec": info.get("instantaneous_ops_per_sec", 0)
}
finally:
await redis_client.close()
class ExternalAPIHealthCheck(BaseHealthCheck):
"""Health check for external APIs."""
def __init__(
self,
name: str,
url: str,
method: str = "GET",
headers: Optional[Dict[str, str]] = None,
expected_status: int = 200,
config: Optional[HealthCheckConfig] = None
):
super().__init__(name, DependencyType.EXTERNAL_API, config)
self.url = url
self.method = method.upper()
self.headers = headers or {}
self.expected_status = expected_status
async def _perform_check(self) -> Dict[str, Any]:
"""Check external API availability."""
async with httpx.AsyncClient() as client:
response = await client.request(
self.method,
self.url,
headers=self.headers,
timeout=self.config.timeout_seconds
)
if response.status_code != self.expected_status:
raise Exception(
f"API returned status {response.status_code}, expected {self.expected_status}"
)
return {
"status_code": response.status_code,
"response_size": len(response.content),
"headers": dict(response.headers),
"url": self.url
}
class LLMServiceHealthCheck(BaseHealthCheck):
"""Health check for LLM services (Groq, OpenAI, etc)."""
def __init__(
self,
name: str,
api_key: str,
base_url: str,
config: Optional[HealthCheckConfig] = None
):
super().__init__(name, DependencyType.LLM_SERVICE, config)
self.api_key = api_key
self.base_url = base_url
async def _perform_check(self) -> Dict[str, Any]:
"""Check LLM service availability with a simple test."""
headers = {
"Authorization": f"Bearer {self.api_key}",
"Content-Type": "application/json"
}
# Simple test prompt
test_payload = {
"model": "llama3-8b-8192", # Groq model
"messages": [
{"role": "user", "content": "Say 'OK' if you can respond."}
],
"max_tokens": 10,
"temperature": 0
}
async with httpx.AsyncClient() as client:
response = await client.post(
f"{self.base_url}/chat/completions",
headers=headers,
json=test_payload,
timeout=self.config.timeout_seconds
)
if response.status_code != 200:
raise Exception(f"LLM service returned status {response.status_code}")
result = response.json()
# Verify response structure
if "choices" not in result or not result["choices"]:
raise Exception("Invalid response structure from LLM service")
return {
"status_code": response.status_code,
"model": result.get("model"),
"usage": result.get("usage", {}),
"response_content": result["choices"][0]["message"]["content"][:50],
"service_name": name
}
class HealthCheckManager:
"""
Manager for coordinating multiple health checks.
Provides centralized health monitoring with dependency tracking
and aggregated health status reporting.
"""
def __init__(self):
self.health_checks: List[BaseHealthCheck] = []
self.last_check_results: Dict[str, HealthCheckResult] = {}
self.check_history: Dict[str, List[HealthCheckResult]] = {}
self.max_history_size = 100
self.logger = get_structured_logger(__name__, component="health_manager")
def register_check(self, health_check: BaseHealthCheck):
"""Register a health check."""
self.health_checks.append(health_check)
self.check_history[health_check.name] = []
self.logger.info(
f"Registered health check: {health_check.name}",
operation="register_health_check",
dependency_type=health_check.dependency_type.value
)
def register_default_checks(self):
"""Register default health checks for the application."""
# Database check
try:
from src.database import get_async_session
db_check = DatabaseHealthCheck(get_async_session)
self.register_check(db_check)
except ImportError:
self.logger.warning("Database session factory not available, skipping DB health check")
# Redis check
if hasattr(settings, 'redis_url') and settings.redis_url:
redis_check = RedisHealthCheck(settings.redis_url)
self.register_check(redis_check)
# Portal da Transparência API check
transparency_check = ExternalAPIHealthCheck(
name="portal_transparencia",
url="https://api.portaldatransparencia.gov.br/api-de-dados/versao",
headers={"accept": "application/json"}
)
self.register_check(transparency_check)
# LLM service check (Groq)
if hasattr(settings, 'groq_api_key') and settings.groq_api_key:
groq_check = LLMServiceHealthCheck(
name="groq_llm",
api_key=settings.groq_api_key,
base_url="https://api.groq.com/openai/v1"
)
self.register_check(groq_check)
async def check_all(self, parallel: bool = True) -> Dict[str, HealthCheckResult]:
"""
Run all registered health checks.
Args:
parallel: Whether to run checks in parallel
Returns:
Dictionary of check results by name
"""
start_time = time.time()
if parallel:
# Run all checks in parallel
tasks = [check.check() for check in self.health_checks]
results = await asyncio.gather(*tasks, return_exceptions=True)
# Process results
check_results = {}
for i, result in enumerate(results):
check_name = self.health_checks[i].name
if isinstance(result, Exception):
# Handle exceptions
check_results[check_name] = HealthCheckResult(
name=check_name,
status=HealthStatus.UNHEALTHY,
dependency_type=self.health_checks[i].dependency_type,
response_time_ms=0,
message="Health check failed with exception",
error=str(result)
)
else:
check_results[check_name] = result
else:
# Run checks sequentially
check_results = {}
for check in self.health_checks:
try:
result = await check.check()
check_results[check.name] = result
except Exception as e:
check_results[check.name] = HealthCheckResult(
name=check.name,
status=HealthStatus.UNHEALTHY,
dependency_type=check.dependency_type,
response_time_ms=0,
message="Health check failed with exception",
error=str(e)
)
# Update history and last results
for name, result in check_results.items():
self.last_check_results[name] = result
if name in self.check_history:
self.check_history[name].append(result)
# Keep only recent history
if len(self.check_history[name]) > self.max_history_size:
self.check_history[name] = self.check_history[name][-self.max_history_size:]
total_time = (time.time() - start_time) * 1000
self.logger.info(
f"Completed health checks for {len(check_results)} dependencies",
operation="health_check_all",
duration_ms=total_time,
parallel=parallel,
results_summary=self._get_status_summary(check_results)
)
return check_results
def get_overall_status(self, results: Optional[Dict[str, HealthCheckResult]] = None) -> HealthStatus:
"""
Get overall system health status.
Args:
results: Optional specific results to evaluate
Returns:
Overall health status
"""
if results is None:
results = self.last_check_results
if not results:
return HealthStatus.UNKNOWN
statuses = [result.status for result in results.values()]
# If any dependency is unhealthy, system is unhealthy
if HealthStatus.UNHEALTHY in statuses:
return HealthStatus.UNHEALTHY
# If any dependency is degraded, system is degraded
if HealthStatus.DEGRADED in statuses:
return HealthStatus.DEGRADED
# All dependencies are healthy
return HealthStatus.HEALTHY
def _get_status_summary(self, results: Dict[str, HealthCheckResult]) -> Dict[str, int]:
"""Get summary of health check statuses."""
summary = {status.value: 0 for status in HealthStatus}
for result in results.values():
summary[result.status.value] += 1
return summary
def get_health_report(self) -> Dict[str, Any]:
"""
Get comprehensive health report.
Returns:
Detailed health report with all dependency information
"""
overall_status = self.get_overall_status()
status_summary = self._get_status_summary(self.last_check_results)
# Calculate availability metrics
total_checks = len(self.last_check_results)
healthy_checks = status_summary.get(HealthStatus.HEALTHY.value, 0)
availability = healthy_checks / total_checks if total_checks > 0 else 0
# Group dependencies by type
dependencies_by_type = {}
for result in self.last_check_results.values():
dep_type = result.dependency_type.value
if dep_type not in dependencies_by_type:
dependencies_by_type[dep_type] = []
dependencies_by_type[dep_type].append(result.to_dict())
return {
"timestamp": datetime.utcnow().isoformat(),
"overall_status": overall_status.value,
"availability_percentage": availability * 100,
"summary": {
"total_dependencies": total_checks,
"healthy_dependencies": healthy_checks,
"degraded_dependencies": status_summary.get(HealthStatus.DEGRADED.value, 0),
"unhealthy_dependencies": status_summary.get(HealthStatus.UNHEALTHY.value, 0),
"status_distribution": status_summary
},
"dependencies_by_type": dependencies_by_type,
"all_dependencies": {
name: result.to_dict()
for name, result in self.last_check_results.items()
}
}
def get_dependency_trends(self, dependency_name: str, hours: int = 24) -> Dict[str, Any]:
"""
Get trend analysis for a specific dependency.
Args:
dependency_name: Name of the dependency
hours: Number of hours to analyze
Returns:
Trend analysis data
"""
if dependency_name not in self.check_history:
return {"error": f"No history found for dependency: {dependency_name}"}
history = self.check_history[dependency_name]
cutoff_time = datetime.utcnow() - timedelta(hours=hours)
# Filter recent history
recent_history = [
result for result in history
if result.timestamp > cutoff_time
]
if not recent_history:
return {"error": f"No recent history for dependency: {dependency_name}"}
# Calculate metrics
response_times = [result.response_time_ms for result in recent_history]
status_counts = {}
for result in recent_history:
status = result.status.value
status_counts[status] = status_counts.get(status, 0) + 1
uptime_percentage = (
status_counts.get(HealthStatus.HEALTHY.value, 0) / len(recent_history) * 100
)
return {
"dependency_name": dependency_name,
"time_window_hours": hours,
"total_checks": len(recent_history),
"uptime_percentage": uptime_percentage,
"avg_response_time_ms": sum(response_times) / len(response_times),
"min_response_time_ms": min(response_times),
"max_response_time_ms": max(response_times),
"status_distribution": status_counts,
"recent_incidents": [
{
"timestamp": result.timestamp.isoformat(),
"status": result.status.value,
"message": result.message,
"response_time_ms": result.response_time_ms
}
for result in recent_history
if result.status != HealthStatus.HEALTHY
]
}
# Global health check manager
health_manager = HealthCheckManager()
async def initialize_health_checks():
"""Initialize default health checks."""
health_manager.register_default_checks()
# Run initial health check
await health_manager.check_all()
logger.info(
"Health check system initialized",
operation="initialize_health_checks",
total_checks=len(health_manager.health_checks)
)