Added files/varnishstat.py

README.md

@@ -81,7 +81,7 @@ While `varnishmon` **doesn't replace a comprehensive monitoring solutions like P
   > Use the `--no-api` flag (or the `api.enabled` setting) to prevent the web interface from starting. You can still collect metrics and store them in the database.
 
 - **Can I customize the metrics collected by `varnishmon`?**
-  > Not directly, as `varnishmon` is designed to collect comprehensive information. However, you can use the `--varnishstat` flag (or the `scraper.varnishstat` setting) to specify a wrapper script that filters the `varnishstat` output (or an inline command if you're up for the quoting challenge).
+  > Not directly, as `varnishmon` is designed to collect comprehensive information. However, you can use the `--varnishstat` flag (or the `scraper.varnishstat` setting) to specify a wrapper script that filters the `varnishstat` output (or an inline command if you're up for the quoting challenge). Check out [this example wrapper script](files/varnishstat.py) for inspiration on how to filter and/or extend the metrics collected.
     > ```bash
     > varnishmon \
     >   --period 15s \

files/varnishmon.yml

@@ -16,6 +16,7 @@ scraper:
   period: 5s
   timeout: 5s
   varnishstat: /mnt/host/files/varnishstat.sh
+  # varnishstat: /usr/local/bin/uv run --quiet --python 3.12 --with psutil==6.1.1 /mnt/host/files/varnishstat.py
 
 api:
   enabled: true

files/varnishstat.py

@@ -0,0 +1,62 @@
+#!/usr/bin/env python3
+# -*- coding: utf-8 -*-
+
+import json
+import re
+import subprocess
+
+import psutil
+
+def varnishstat():
+    result = subprocess.run(
+        ['/usr/bin/varnishstat', '-1', '-j'],
+        capture_output=True,
+        text=True)
+    assert result.returncode == 0, f'Command failed with exit code {result.returncode}'
+    return json.loads(result.stdout)
+
+def main():
+    # Get 'varnishstat' metrics.
+    metrics = varnishstat()
+
+    # Filter out some metrics we prefer to ignore.
+    blacklist = re.compile(r'^(?:LCK|MEMPOOL)[.]')
+    metrics = {k: v for k, v in metrics.items() if not blacklist.match(k)}
+
+    # Add some additional ad-hoc metrics following the 'varnishstat' output
+    # format.
+    for k, v in psutil.cpu_times(percpu=False)._asdict().items():
+        metrics[f'CPU.time.{k}'] = {
+            'description': f"psutil.cpu_times(percpu=False)._asdict()['{k}'] * 1000",
+            'flag': 'c',
+            'format': 'i',
+            'value': int(v * 1000),
+        }
+    for k, v in psutil.virtual_memory()._asdict().items():
+        metrics[f'MEMORY.{k}'] = {
+            'description': f"psutil.virtual_memory()._asdict()['{k}']",
+            'flag': 'g',
+            'format': 'i' if k in ['percent'] else 'B',
+            'value': int(v),
+        }
+    for k, v in psutil.swap_memory()._asdict().items():
+        metrics[f'SWAP.{k}'] = {
+            'description': f"psutil.swap_memory()._asdict()['{k}']",
+            'flag': 'g' if k in ['total', 'used', 'free', 'percent'] else 'c',
+            'format': 'i' if k in ['percent'] else 'B',
+            'value': int(v),
+        }
+    for nic, counters in psutil.net_io_counters(pernic=True).items():
+        for k, v in counters._asdict().items():
+            metrics[f'NET.{nic}.{k}'] = {
+                'description': f"psutil.net_io_counters(pernic=True)['{nic}']._asdict()['{k}']",
+                'flag': 'c',
+                'format': 'B' if 'bytes' in k else 'i',
+                'value': int(v),
+            }
+
+    # Print metrics as JSON.
+    print(json.dumps(metrics))
+
+if __name__ == '__main__':
+    main()

files/varnishstat.sh

@@ -25,8 +25,6 @@
 #   - Beware of JSON numbers being limited to float64 & possible precision loss.
 #
 
-/usr/bin/varnishstat -1 -j && exit 0
-
 cat <<EOF
 {
   "timestamp": "2024-01-01T13:00:00",