ghaymah-exam-Moataz-Gamal-Hassan-SRE/q5-mithal-monitor/monitor.py

#!/usr/bin/env python3
"""
Monitoring script for mithal.space
Collects every 60 seconds:
  - Latency: HTTP response time for the homepage
  - Uptime: whether the site responds with a healthy status code
  - SSL: certificate validity and days remaining until expiry
  - DNS: DNS resolution time
  - Search response: time to get a response from the search endpoint

Storage:
  - dashboard/data/metrics_log.csv     -> full historical log (append-only)
  - dashboard/data/metrics_history.json -> rolling window (last 24h) used by the dashboard

Only standard library is used - no extra dependencies to install.
"""

import csv
import json
import socket
import ssl
import time
import urllib.request
import urllib.error
from datetime import datetime, timezone
from pathlib import Path

# ---------------------------------------------------------------------------
# Configuration
# ---------------------------------------------------------------------------
TARGET_HOST = "mithal.space"
TARGET_URL = "https://mithal.space/"
SEARCH_URL = "https://mithal.space/search?q=test"   # adjust if the real search path differs
CHECK_INTERVAL_SECONDS = 60
REQUEST_TIMEOUT_SECONDS = 10
ROLLING_WINDOW_MINUTES = 24 * 60   # keep last 24 hours (1 check/minute)

DATA_DIR = Path(__file__).parent / "dashboard" / "data"
CSV_LOG_PATH = DATA_DIR / "metrics_log.csv"
HISTORY_JSON_PATH = DATA_DIR / "metrics_history.json"

CSV_FIELDS = [
    "timestamp", "up", "status_code", "latency_ms",
    "dns_ms", "ssl_valid", "ssl_days_remaining", "ssl_expiry",
    "search_status", "search_latency_ms", "error",
]


def now_iso():
    return datetime.now(timezone.utc).isoformat()


def check_dns(host):
    """Measure DNS resolution time in milliseconds."""
    start = time.perf_counter()
    try:
        socket.getaddrinfo(host, 443)
        elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
        return elapsed_ms, None
    except socket.gaierror as exc:
        elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
        return elapsed_ms, str(exc)


def check_ssl(host, port=443):
    """Connect to the host and read certificate expiry info."""
    context = ssl.create_default_context()
    try:
        with socket.create_connection((host, port), timeout=REQUEST_TIMEOUT_SECONDS) as sock:
            with context.wrap_socket(sock, server_hostname=host) as ssock:
                cert = ssock.getpeercert()
        not_after = cert.get("notAfter")
        expiry_dt = datetime.strptime(not_after, "%b %d %H:%M:%S %Y %Z").replace(tzinfo=timezone.utc)
        days_remaining = (expiry_dt - datetime.now(timezone.utc)).days
        return {
            "ssl_valid": True,
            "ssl_days_remaining": days_remaining,
            "ssl_expiry": expiry_dt.isoformat(),
        }, None
    except Exception as exc:
        return {
            "ssl_valid": False,
            "ssl_days_remaining": None,
            "ssl_expiry": None,
        }, str(exc)


def check_http(url):
    """Measure HTTP response time and status code."""
    start = time.perf_counter()
    try:
        req = urllib.request.Request(url, headers={"User-Agent": "mithal-monitor/1.0"})
        with urllib.request.urlopen(req, timeout=REQUEST_TIMEOUT_SECONDS) as resp:
            elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
            return resp.getcode(), elapsed_ms, None
    except urllib.error.HTTPError as exc:
        elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
        return exc.code, elapsed_ms, None
    except Exception as exc:
        elapsed_ms = round((time.perf_counter() - start) * 1000, 2)
        return None, elapsed_ms, str(exc)


def run_check():
    record = {"timestamp": now_iso(), "error": None}

    # 1) DNS resolution time
    dns_ms, dns_err = check_dns(TARGET_HOST)
    record["dns_ms"] = dns_ms

    # 2) SSL certificate status
    ssl_info, ssl_err = check_ssl(TARGET_HOST)
    record.update(ssl_info)

    # 3) HTTP latency + uptime (status code)
    status_code, latency_ms, http_err = check_http(TARGET_URL)
    record["status_code"] = status_code
    record["latency_ms"] = latency_ms
    record["up"] = status_code is not None and 200 <= status_code < 400

    # 4) Search response time
    search_status, search_latency_ms, search_err = check_http(SEARCH_URL)
    record["search_status"] = search_status
    record["search_latency_ms"] = search_latency_ms

    errors = [e for e in (dns_err, ssl_err, http_err, search_err) if e]
    if errors:
        record["error"] = " | ".join(errors)

    return record


def append_csv(record):
    DATA_DIR.mkdir(parents=True, exist_ok=True)
    file_exists = CSV_LOG_PATH.exists()
    with CSV_LOG_PATH.open("a", newline="", encoding="utf-8") as f:
        writer = csv.DictWriter(f, fieldnames=CSV_FIELDS)
        if not file_exists:
            writer.writeheader()
        writer.writerow(record)


def update_history_json(record):
    DATA_DIR.mkdir(parents=True, exist_ok=True)
    history = []
    if HISTORY_JSON_PATH.exists():
        try:
            history = json.loads(HISTORY_JSON_PATH.read_text(encoding="utf-8"))
        except json.JSONDecodeError:
            history = []
    history.append(record)
    history = history[-ROLLING_WINDOW_MINUTES:]
    HISTORY_JSON_PATH.write_text(json.dumps(history, indent=2, ensure_ascii=False), encoding="utf-8")


def main():
    print(f"[{now_iso()}] Starting monitor for {TARGET_URL} (every {CHECK_INTERVAL_SECONDS}s)")
    while True:
        record = run_check()
        append_csv(record)
        update_history_json(record)

        status_label = "UP" if record["up"] else "DOWN"
        print(
            f"[{record['timestamp']}] {status_label} "
            f"status={record['status_code']} latency={record['latency_ms']}ms "
            f"dns={record['dns_ms']}ms ssl_days={record['ssl_days_remaining']} "
            f"search={record['search_latency_ms']}ms"
        )
        if record["error"]:
            print(f"  -> issues: {record['error']}")

        time.sleep(CHECK_INTERVAL_SECONDS)


if __name__ == "__main__":
    main()