ghaymah-exam-noran-sre/q5-mithal-monitor/monitor.py

#!/usr/bin/env python3
"""
Monitoring Script for mithal.space
Collects: latency, uptime, SSL, DNS, search response
"""

import requests
import json
import time
import socket
import ssl
import datetime
import os
from urllib.parse import urlparse

# Configuration
TARGET_URL = "https://mithal.space"
SEARCH_QUERY = "?q=test"          # adjust if search endpoint differs
DATA_FILE = "monitoring-data.json"
MAX_ENTRIES = 1440                # 24 hours * 60 minutes
TIMEOUT = 10

def check_ssl_certificate(hostname, port=443):
    """Check SSL certificate expiry date."""
    try:
        context = ssl.create_default_context()
        with socket.create_connection((hostname, port), timeout=TIMEOUT) as sock:
            with context.wrap_socket(sock, server_hostname=hostname) as ssock:
                cert = ssock.getpeercert()
                expiry_date = datetime.datetime.strptime(cert['notAfter'], '%b %d %H:%M:%S %Y %Z')
                days_remaining = (expiry_date - datetime.datetime.utcnow()).days
                return {
                    "expiry_date": expiry_date.isoformat(),
                    "days_remaining": days_remaining,
                    "valid": days_remaining > 0
                }
    except Exception as e:
        return {
            "expiry_date": None,
            "days_remaining": None,
            "valid": False,
            "error": str(e)
        }

def resolve_dns(hostname):
    """Measure DNS resolution time."""
    try:
        start = time.time()
        socket.gethostbyname(hostname)
        dns_time = round((time.time() - start) * 1000, 2)
        return dns_time
    except Exception:
        return None

def check_latency_and_uptime(url):
    """Check HTTP response time and status code."""
    try:
        start = time.time()
        response = requests.get(url, timeout=TIMEOUT)
        latency = round((time.time() - start) * 1000, 2)
        status_code = response.status_code
        is_up = 200 <= status_code < 400
        return {
            "latency": latency,
            "status_code": status_code,
            "is_up": is_up,
            "error": None
        }
    except requests.RequestException as e:
        return {
            "latency": None,
            "status_code": None,
            "is_up": False,
            "error": str(e)
        }

def check_search_response(url, query="?q=test"):
    """Measure search endpoint response time."""
    try:
        search_url = f"{url}{query}"
        start = time.time()
        response = requests.get(search_url, timeout=TIMEOUT)
        search_latency = round((time.time() - start) * 1000, 2)
        return {
            "search_latency": search_latency,
            "search_status": response.status_code,
            "search_error": None
        }
    except requests.RequestException as e:
        return {
            "search_latency": None,
            "search_status": None,
            "search_error": str(e)
        }

def collect_metrics():
    """Collect all metrics for mithal.space."""
    parsed = urlparse(TARGET_URL)
    hostname = parsed.hostname or "mithal.space"
    
    dns_time = resolve_dns(hostname)
    ssl_info = check_ssl_certificate(hostname)
    health = check_latency_and_uptime(TARGET_URL)
    search = check_search_response(TARGET_URL, SEARCH_QUERY)
    
    metrics = {
        "timestamp": datetime.datetime.utcnow().isoformat() + "Z",
        "url": TARGET_URL,
        "dns_ms": dns_time,
        "ssl_valid": ssl_info.get("valid", False),
        "ssl_days_remaining": ssl_info.get("days_remaining"),
        "ssl_expiry_date": ssl_info.get("expiry_date"),
        "latency_ms": health.get("latency"),
        "status_code": health.get("status_code"),
        "is_up": health.get("is_up", False),
        "error": health.get("error"),
        "search_latency_ms": search.get("search_latency"),
        "search_status_code": search.get("search_status"),
        "search_error": search.get("search_error")
    }
    return metrics

def save_metrics(metrics, filename=DATA_FILE, max_entries=MAX_ENTRIES):
    """Save metrics to JSON file with size limit."""
    try:
        if os.path.exists(filename):
            with open(filename, 'r') as f:
                data = json.load(f)
        else:
            data = []
        data.append(metrics)
        if len(data) > max_entries:
            data = data[-max_entries:]
        with open(filename, 'w') as f:
            json.dump(data, f, indent=2)
        return True
    except Exception as e:
        print(f"Error saving data: {e}")
        return False

def run_monitor():
    """Main monitoring loop."""
    print(f"🚀 Starting monitoring for {TARGET_URL}")
    print(f"📊 Data will be saved to {DATA_FILE}")
    print("=" * 50)
    
    while True:
        try:
            print(f"[{datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')}] Collecting metrics...")
            metrics = collect_metrics()
            
            status = "✅ UP" if metrics['is_up'] else "❌ DOWN"
            print(f"  Status: {status}")
            print(f"  Latency: {metrics['latency_ms']} ms")
            print(f"  DNS: {metrics['dns_ms']} ms")
            print(f"  SSL: {metrics['ssl_days_remaining']} days remaining")
            print(f"  Search: {metrics['search_latency_ms']} ms")
            
            if save_metrics(metrics):
                print("  ✅ Data saved")
            else:
                print("  ❌ Failed to save data")
            print("-" * 50)
            
        except KeyboardInterrupt:
            print("\n🛑 Monitoring stopped by user")
            break
        except Exception as e:
            print(f"❌ Error in monitoring loop: {e}")
        
        time.sleep(60)

if __name__ == "__main__":
    run_monitor()