Real-time monitoring dashboard for Proxmox VE, PBS, Docker, Kubernetes, TrueNAS and vSphere. Self-hosted, with alerts for failed backups, full disks, unhealthy containers and offline nodes.
Go aialertsdashboarddockerdocker-monitoringgohomelabinfrastructure-monitoringkubernetesmonitoringmspproxmoxproxmox-backup-serverproxmox-mail-gatewayproxmox-veself-hostedtruenastypescriptvspherewebhooks
#!/usr/bin/env python3
"""Run one audit with an owned deadline and process group, preserving its bytes."""
from __future__ import annotations
import os
import signal
import subprocess
import sys
import time
from pathlib import Path
def signal_group(process: subprocess.Popen, signum: int) -> None:
# start_new_session makes this the audit's group, never the invoking
# shell's or another job's. The leader may have exited with children left.
try:
os.killpg(process.pid, signum)
except ProcessLookupError:
pass
def run_audit(limit: float, output: Path, command: list[str]) -> int:
process = None
status = None
timed_out = False
interrupted = [None]
previous_handlers = {}
def interrupt(signum: int, _frame: object) -> None:
# Do not raise between Popen creating the child and returning its
# handle: cleanup must have the actual group identity before it acts.
interrupted[0] = signum
try:
for signum in (signal.SIGTERM, signal.SIGINT, signal.SIGHUP):
previous_handlers[signum] = signal.signal(signum, interrupt)
with output.open("wb") as report:
if interrupted[0] is None:
process = subprocess.Popen(
command, stdout=report, stderr=subprocess.DEVNULL,
start_new_session=True,
)
deadline = time.monotonic() + limit
while interrupted[0] is None:
remaining = deadline - time.monotonic()
if remaining <= 0:
timed_out = True
break
try:
status = process.wait(timeout=min(0.1, remaining))
break
except subprocess.TimeoutExpired:
pass
if timed_out:
signal_group(process, signal.SIGTERM)
try:
status = process.wait(timeout=2)
except subprocess.TimeoutExpired:
pass
except OSError as exc:
print(f"npm audit process could not start: {exc.strerror}", file=sys.stderr)
status = 127
finally:
# A second cancellation must not interrupt cleanup. KILL also owns
# TERM-resistant descendants after a timeout or an already exited
# leader. No sleeper/watchdog process is needed on the success path.
for signum in previous_handlers:
signal.signal(signum, signal.SIG_IGN)
try:
if process is not None:
signal_group(process, signal.SIGKILL)
observed_status = process.wait()
if status is None:
status = observed_status
finally:
for signum, handler in previous_handlers.items():
signal.signal(signum, handler)
if interrupted[0] is not None:
return 128 + interrupted[0]
# A process that exits zero in its TERM handler still missed the deadline.
# Its zero summary must not become a clean audit.
if timed_out and status >= 0:
status = -signal.SIGTERM
return status if status >= 0 else 128 - status
def main() -> int:
if len(sys.argv) < 4:
print("Usage: npm_audit_process.py seconds report command [args...]", file=sys.stderr)
return 2
try:
limit = float(sys.argv[1])
if not 0 < limit < float("inf"):
raise ValueError
except ValueError:
print("npm audit process needs a finite positive deadline", file=sys.stderr)
return 2
return run_audit(limit, Path(sys.argv[2]), sys.argv[3:])
if __name__ == "__main__":
sys.exit(main())