Add system and certificate telemetry collectors

collectors/certificates.py

@@ -0,0 +1,61 @@
+#!/usr/bin/env python3
+
+import subprocess
+from datetime import datetime, timezone
+from pathlib import Path
+
+CERT_ROOT = Path(
+    "/opt/docker/caddy/data/caddy/certificates/"
+    "acme-v02.api.letsencrypt.org-directory"
+)
+
+
+def collect():
+    certificates = []
+
+    for cert_file in sorted(CERT_ROOT.glob("*/*.crt")):
+        result = subprocess.run(
+            ["openssl", "x509", "-in", str(cert_file), "-noout", "-enddate"],
+            capture_output=True,
+            text=True,
+        )
+
+        if result.returncode != 0:
+            continue
+
+        expires_text = result.stdout.strip().removeprefix("notAfter=")
+        expires = datetime.strptime(
+            expires_text,
+            "%b %d %H:%M:%S %Y %Z",
+        ).replace(tzinfo=timezone.utc)
+
+        days_remaining = (expires - datetime.now(timezone.utc)).days
+
+        certificates.append({
+            "name": cert_file.parent.name,
+            "days_remaining": days_remaining,
+            "expires": expires.isoformat(),
+        })
+
+    if not certificates:
+        return {
+            "state": "offline",
+            "summary": "No certificates",
+            "detail": "Caddy certificate files not found",
+        }
+
+    lowest = min(item["days_remaining"] for item in certificates)
+
+    state = "online"
+    if lowest < 30:
+        state = "warning"
+    if lowest < 7:
+        state = "offline"
+
+    return {
+        "state": state,
+        "summary": f"{len(certificates)} certificates",
+        "detail": f"{lowest} days minimum",
+        "minimum_days": lowest,
+        "certificates": certificates,
+    }

collectors/system.py

@@ -0,0 +1,51 @@
+#!/usr/bin/env python3
+
+import os
+import shutil
+
+
+def collect():
+    load_1, _, _ = os.getloadavg()
+
+    memory_total = 0
+    memory_available = 0
+
+    with open("/proc/meminfo", "r", encoding="utf-8") as handle:
+        for line in handle:
+            key, value = line.split(":", 1)
+            kb = int(value.strip().split()[0])
+
+            if key == "MemTotal":
+                memory_total = kb
+            elif key == "MemAvailable":
+                memory_available = kb
+
+    memory_used_pct = round(
+        (1 - memory_available / memory_total) * 100
+    ) if memory_total else 0
+
+    disk = shutil.disk_usage("/")
+    disk_used_pct = round(disk.used / disk.total * 100)
+
+    with open("/proc/uptime", "r", encoding="utf-8") as handle:
+        uptime_seconds = int(float(handle.read().split()[0]))
+
+    uptime_days = uptime_seconds // 86400
+
+    state = "online"
+
+    if memory_used_pct >= 85 or disk_used_pct >= 85:
+        state = "warning"
+
+    if memory_used_pct >= 95 or disk_used_pct >= 95:
+        state = "offline"
+
+    return {
+        "state": state,
+        "summary": f"{memory_used_pct}% memory",
+        "detail": f"{disk_used_pct}% disk · {uptime_days} days uptime",
+        "load_1": round(load_1, 2),
+        "memory_used_pct": memory_used_pct,
+        "disk_used_pct": disk_used_pct,
+        "uptime_days": uptime_days,
+    }