Files
voidea/tools/metrics_service.py
T

127 lines
3.6 KiB
Python

"""Simple system metrics collector for VoIdea VPS.
Runs as a periodic task (Celery or cron) and logs:
- CPU load
- Memory usage
- Disk usage
- PostgreSQL connection count
"""
import json
import logging
import os
import platform
import shutil
import subprocess
import sys
from datetime import datetime, timezone
logger = logging.getLogger("voidea.metrics")
def get_cpu_load() -> float:
"""Return CPU load average (1 min) or 0.0."""
try:
if platform.system() == "Linux":
load1, _, _ = os.getloadavg()
return round(load1, 2)
return 0.0
except OSError:
return 0.0
def get_memory_usage() -> dict:
"""Return memory usage in MB."""
try:
import psutil
mem = psutil.virtual_memory()
return {
"total_mb": round(mem.total / 1024 / 1024, 1),
"available_mb": round(mem.available / 1024 / 1024, 1),
"percent_used": mem.percent,
}
except ImportError:
return {"error": "psutil not installed"}
def get_disk_usage(path: str = "/") -> dict:
"""Return disk usage for the given path."""
try:
usage = shutil.disk_usage(path)
return {
"total_gb": round(usage.total / 1024**3, 1),
"used_gb": round(usage.used / 1024**3, 1),
"free_gb": round(usage.free / 1024**3, 1),
"percent_used": round(usage.used / usage.total * 100, 1),
}
except OSError:
return {"error": f"cannot access {path}"}
def get_pg_connections(db_url: str | None = None) -> int:
"""Return active PostgreSQL connection count."""
url = db_url or os.environ.get("DATABASE_URL", "")
if not url:
return -1
try:
parts = url.replace("postgresql://", "").replace("postgres://", "").split("@")
user_pass = parts[0].split(":")
host_db = parts[1].split("/")
host_port = host_db[0].split(":")
config = {
"host": host_port[0],
"port": host_port[1] if len(host_port) > 1 else "5432",
"user": user_pass[0],
"dbname": host_db[1].split("?")[0],
}
env = os.environ.copy()
env["PGPASSWORD"] = user_pass[1] if len(user_pass) > 1 else ""
result = subprocess.run(
[
"psql",
"-h", config["host"],
"-p", config["port"],
"-U", config["user"],
"-d", config["dbname"],
"-tA",
"-c", "SELECT count(*) FROM pg_stat_activity;",
],
env=env,
capture_output=True,
text=True,
timeout=5,
)
return int(result.stdout.strip())
except (ValueError, subprocess.TimeoutExpired, OSError):
return -1
def collect() -> dict:
"""Collect all metrics into a single dict."""
return {
"timestamp": datetime.now(timezone.utc).isoformat(),
"hostname": platform.node(),
"cpu_load_1min": get_cpu_load(),
"memory": get_memory_usage(),
"disk": get_disk_usage("/"),
"pg_connections": get_pg_connections(),
}
def main():
logging.basicConfig(level=logging.INFO, format="%(asctime)s [%(levelname)s] %(message)s")
metrics = collect()
log_line = json.dumps(metrics, ensure_ascii=False)
logger.info("Metrics: %s", log_line)
# Optionally write to a file for prometheus/node_exporter
metrics_file = os.environ.get("METRICS_FILE", "")
if metrics_file:
os.makedirs(os.path.dirname(metrics_file), exist_ok=True)
with open(metrics_file, "w", encoding="utf-8") as f:
f.write(log_line + "\n")
if __name__ == "__main__":
main()