From 0a7b811605d059eee4a8d43a6fb09b491d2d9ef2 Mon Sep 17 00:00:00 2001
From: Martin Kompf <mkompf@chronos>
Date: Mon, 1 Aug 2016 17:53:33 +0200
Subject: [PATCH] Leak detection

---
 leakdetect.py | 85 +++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 85 insertions(+)
 create mode 100755 leakdetect.py

diff --git a/leakdetect.py b/leakdetect.py
new file mode 100755
index 0000000..ea0a572
--- /dev/null
+++ b/leakdetect.py
@@ -0,0 +1,85 @@
+#!/usr/bin/env python
+
+# Script to detect leakages in water consumption by reading the timestamps
+# and counter values from a database.
+# The script assumes that during a period of some hours of a day should
+# be absolute no water consumption.
+# The script detects the longest pause of water consumption. If its duration
+# is below a given threshold, then a warning message is printed.
+
+import os
+import csv
+import rrdtool as rrd
+from datetime import datetime
+from math import *
+
+# Path to RRD with counter values
+count_rrd = "{0}/water.rrd".format(os.path.dirname(os.path.abspath(__file__)))
+
+# Alternative: Path to output of rrdtool fetch
+count_fetch = "{0}/water.fetch.txt".format(os.path.dirname(os.path.abspath(__file__)))
+
+# Length of the minimum duration without water consumption in seconds 
+min_pause = 3 * 60 * 60 # 3 hours
+
+# Verbose output
+verbose = True
+
+# Read the rrd specified in count_rrd.
+# Returns array of (timestamp, counter).
+def read_rrd():
+  result = []
+  ((start, stop, step), head, data) = rrd.fetch(count_rrd, 'LAST', '-s', 'now - 1 day', '-e', 'now')
+  t = start
+  for row in data:
+    if row[0]:
+      result.append((t, row[0])) 
+    t += step
+  return result
+
+# Read the text file specified in count_fetch.
+# The text file is the output of command 'rrdtool fetch'.
+# Returns array of (timestamp, counter).
+def read_fetch_output():
+  result = []
+  with open(count_fetch, 'r') as f:
+    reader = csv.reader(f, delimiter=' ')
+    for row in reader:
+      if len(row) == 3:
+        value = int(row[0][:-1]), float(row[1])  
+        if not isnan(value[1]):
+          result.append(value) 
+  return result
+
+# Detect all pauses which last longer than min_pause.
+# Param data: array of (timestamp, counter)
+# Returns array of dictionaries {start, duration}.
+def detect_pauses(data):
+  result = []
+  (start_time, start_counter) = data[0]
+  (end_time, end_counter) = data[-1]
+  data.append((end_time + 1, end_counter + 1)) # ensure that counter increments
+  for (ts, counter) in data:
+    if counter > start_counter:
+      duration = ts - start_time
+      if duration > min_pause:
+        result.append({'start' : start_time, 'duration' : duration})
+      start_time = ts
+      start_counter = counter
+  return result
+
+# MAIN
+def main():
+  #counter = read_fetch_output()
+  counter = read_rrd()
+  pauses = detect_pauses(counter)
+  if verbose:
+    for p in pauses:
+      print("Pause starting at {0:%Y-%m-%d %H:%M:%S}: {1:.1f} hours"
+        .format(datetime.fromtimestamp(p['start']), p['duration']/3600.0))
+  if len(pauses) == 0:
+    print("Possible leak detected! There is no break of at least {0} hours."
+      .format(min_pause/3600.0))
+
+if __name__ == '__main__':
+  main()