WIP implementing flapping avoidance

healthcheck/healthcheck.cfg.example

@@ -80,6 +80,14 @@ NOTIFY_ALARM_END=TRUE
 # success result, error result, command failure. Then paste the command
 # and regex in this config, enable the check and run to verify is working.
 
+#### AVOID FLAPPING ####
+# For values slowly changing, like a server cabinet temperature, a lot of start/end alarm notifications
+# may be generated if the value is near the limit (e.g. alarm set to 30, temperature jumping between 30.1
+# and 29.9). To avoid this, a "grace" delta can be set with:
+# DELTA=0.5
+# In the previous example, this make the alarm fire at 30.5 and stop at 29.5.
+# If not specified, the delta is 0.
+
 
 [system_load_1min]
 # The system load average in the last minute

healthcheck/healthcheck.py

@@ -169,13 +169,20 @@ class Main:
 
 		# Compare detected value with equal, not equal, more than and less values
 		logging.info('detected {}'.format(detectedValue))
+
+		# String comparison
 		if config.alarm_string_equal and (detectedValue == config.alarm_string_equal):
 			return 'value is "{}"'.format(detectedValue)
 		if config.alarm_string_not_equal and (detectedValue != config.alarm_string_not_equal):
 			return 'value is "{}", but should be "{}"'.format(detectedValue, config.alarm_string_not_equal)
-		if config.alarm_value_equal and (locale.atof(detectedValue) == float(config.alarm_value_equal)):
+
+		# Numeric comparison
+		numeric_value = locale.atof(detectedValue)
+		delta_upper = numeric_value + config.delta
+		delta_lower = numeric_value - config.delta
+		if config.alarm_value_equal and (delta_lower < float(config.alarm_value_equal) or delta_upper > float(config.alarm_value_equal)):
 			return 'value is {}'.format(detectedValue)
-		if config.alarm_value_not_equal and (locale.atof(detectedValue) != float(config.alarm_value_not_equal)):
+		if config.alarm_value_not_equal and (delta_lower > float(config.alarm_value_not_equal) or delta_upper < float(config.alarm_value_not_equal)):
 			return 'value is {}, but should be {}'.format(detectedValue, config.alarm_value_not_equal)
 		if config.alarm_value_more_than and locale.atof(detectedValue) > float(config.alarm_value_more_than):
 			return 'value is {}, but should not exceed {}'.format(locale.atof(detectedValue), config.alarm_value_more_than)