Updated readme, enabled by default universal checks

.gitignore

@@ -1 +1,2 @@
 *.cfg
+healthcheck/healthcheck-virtualenv

README.md

@@ -4,6 +4,8 @@ Every utility is in a folder with its relevant configuration and is completely s
 
 ## HEALTHCHECK
 A simple server health check.
+Allows to keep under control the machine vitals (cpu usage, raid status, thermals...) and alter the sysadmin in case of anomalies.
+
 Sends an email and/or executes a command in case of alarm (high temperature, RAID disk failed etc...).
 As an example, the command may be a ntfy call to obtain a notification on a mobile phone or desktop computer.
 Meant to be run with a cron (see healthcheck.cron.example).

healthcheck/README.md

@@ -1,5 +1,7 @@
 # HEALTHCHECK
 A simple server health check.
+Allows to keep under control the machine vitals (cpu usage, raid status, thermals...) and alter the sysadmin in case of anomalies.
+
 Sends an email and/or executes a command in case of alarm.
 As an example, the command may be a ntfy call to obtain a notification on a mobile phone or desktop computer.
 Meant to be run with a cron (see healthcheck.cron.example).
@@ -40,6 +42,10 @@ Copy the script and the config file into the system to check:
 cp healthcheck.py /usr/local/bin/healthcheck.py
 cp healthcheck.cfg.example /usr/local/etc/healthcheck.cfg
 ```
+Make the script executable:
+```
+chmod +x /usr/local/bin/healthcheck.py
+```
 Edit `/usr/local/etc/healthcheck.cfg` enabling the checks you need and configuring email settings.
 Run `/usr/local/bin/healthcheck.py /usr/local/etc/healthcheck.cfg` to check it is working. If needed, change the config to make a check fail and see if the notification mail is delivered. If you need to do some testing without spamming emails, run with the parameter `--dry-run`.
 Now copy the cron file:

healthcheck/healthcheck.cfg.example

@@ -1,3 +1,5 @@
+# The DEFAULT section contains the global configuration applied to all checks.
+# You can re-define this variables in a check to override the global one.
 [DEFAULT]
 
 #### EMAIL NOTIFICATIONS ####
@@ -39,10 +41,27 @@ MAILTO=root@localhost, user@localhost
 #ALARM_COMMAND=curl -H "%%CHECKNAME%% alarm on %%HOSTNAME%%" -d "%%ERROR%% on %%DATETIME%%" ntfy.sh/my-unique-topic-name
 
 
+#### NOTIFICATION POLICY ###
+# Defines when to send the email and/or execute ALARM_COMMAND. Useful to avoid email flooding.
+# Possible values:
+# EVERY_RUN In case of alarm, sends a mail every time the script is run
+# START Sends a mail only when an alarm starts
+# ONCE_IN_MINUTES In case of alarm, resends a mail only if NOTIFY_MINUTES has passed
+NOTIFY=EVERY_RUN
+
+# Used only if NOTIFY=ONCE_IN_MINUTES. A mail is sent only if NOTIFY_MINUTES has passed from the previous one
+NOTIFY_MINUTES=60
+
+# Sends a mail when the alarm has ended
+NOTIFY_ALARM_END=TRUE
+
+
 #### HEALTH CHECKS ####
 # Every health check is based on a command being executed, its result being parsed with a regexp
 # to extract (as a single group) the numeric or string value, and the value being compared with
 # a configured value. This checks are ready to be used, just enable the ones you need.
+#
+# CUSTOM CHECKS:
 # You can add your own custom check declaring another section like this:
 #
 # [my_custom_check_name]
@@ -55,28 +74,37 @@ MAILTO=root@localhost, user@localhost
 # ALARM_VALUE_LESS_THAN=12
 # COMMAND=/my/custom/binary --with parameters
 # REGEXP=my regex to parse (awesome|disappointing) command output
+#
+# First test your custom command executing it in the command line
+# Take the text output and write a regex to match it. Check every case:
+# success result, error result, command failure. Then paste the command
+# and regex in this config, enable the check and run to verify is working.
+
 
 [system_load_1min]
 # The system load average in the last minute
-DISABLED=True
+DISABLED=False
 ALARM_VALUE_MORE_THAN=1.0
 COMMAND=uptime
 REGEXP=.*load average: (\d+[,.]\d+), \d+[,.]\d+, \d+[,.]\d+
 
+
 [system_load_5min]
 # The system load average in the last 5 minutes
-DISABLED=True
+DISABLED=False
 ALARM_VALUE_MORE_THAN=1.0
 COMMAND=uptime
 REGEXP=.*load average: \d+[,.]\d+, (\d+[,.]\d+), \d+[,.]\d+
 
+
 [system_load_15min]
 # The system load average in the last 15 minutes
-DISABLED=True
+DISABLED=False
 ALARM_VALUE_MORE_THAN=1.0
 COMMAND=uptime
 REGEXP=.*load average: \d+[,.]\d+, \d+[,.]\d+, (\d+[,.]\d+)
 
+
 [used_disk_space]
 # Used disk space (in percent, i.e. ALARM_VALUE_MORE_THAN=75 -> alarm if disk is more than 75% full)
 DISABLED=True
@@ -84,6 +112,7 @@ ALARM_VALUE_MORE_THAN=75
 COMMAND=df -h /dev/sda1
 REGEXP=(\d{1,3})%
 
+
 [raid_status]
 # Issues an alarm when the raid is corrupted
 # Checks this part of the /proc/mdstat file:
@@ -95,6 +124,7 @@ ALARM_STRING_NOT_EQUAL=UU
 COMMAND=cat /proc/mdstat
 REGEXP=.*\] \[([U_]+)\]\n
 
+
 [battery_level]
 # Issues an alarm when battery is discharging below a certain level (long blackout, pulled power cord...)
 # For laptops used as servers, apparently common among the self hosters. Requires acpi package installed.
@@ -104,6 +134,7 @@ COMMAND=acpi -b
 REGEXP=Battery \d: .*, (\d{1,3})%
 ALARM_VALUE_LESS_THAN=90
 
+
 [laptop_charger_disconnected]
 # Issues an alarm when laptop charger is disconnected
 # For laptops used as servers, apparently common among the self hosters. Requires acpi package installed.
@@ -112,20 +143,14 @@ COMMAND=acpi -a
 REGEXP=Adapter \d: (.+)
 ALARM_STRING_EQUAL=off-line
 
-[free_ram]
-# Free ram in %
-# Shows another approach: does all the computation in the command and picks up
-# all the output (by not declaring a regexp).
-DISABLED=True
-COMMAND=free | grep Mem | awk '{print int($4/$2 * 100.0)}'
-ALARM_VALUE_LESS_THAN=20
 
 [available_ram]
-# Like Free ram, but shows available instead of free. You may want to use this if you use a memcache.
+# Shows available ram in %.
-DISABLED=True
+DISABLED=False
 COMMAND=free | grep Mem | awk '{print int($7/$2 * 100.0)}'
 ALARM_VALUE_LESS_THAN=20
 
+
 [cpu_temperature]
 # CPU Temperature alarm: requires lm-sensors installed and configured (check your distribution's guide)
 # The regexp must be adapted to your configuration: run `sensors` in the command line
@@ -136,6 +161,7 @@ ALARM_VALUE_MORE_THAN=80
 COMMAND=sensors
 REGEXP=Core 0: +\+?(-?\d{1,3}).\d°[CF]
 
+
 [fan_speed]
 # Fan speed alarm: requires lm-sensors installed and configured (check your distribution's guide)
 # The regexp must be adapted to your configuration: run `sensors` in the command line
@@ -144,3 +170,38 @@ DISABLED=True
 ALARM_VALUE_LESS_THAN=300
 COMMAND=sensors
 REGEXP=cpu_fan: +(\d) RPM
+
+
+[host_reachability]
+# Check if a remote host is alive with Ping. You can replace the ip with a domain name (e.g. COMMAND=ping debian.org -c 1)
+#
+# Shows another approach: uses the return value to print a string. Leverages ping's ability to return different error codes:
+# 0 = success
+# 1 = the host is unreachable
+# 2 = an error has occurred (and will be logged to stderr)
+# We are throwing away stdout and replacing it with a custom text.
+# If there is a different text (the stderr), something bad happened, and it will be reported in the mail.
+DISABLED=True
+ALARM_STRING_NOT_EQUAL=Online
+COMMAND=ping 192.168.1.123 -c 1 > /dev/null && echo "Online" || echo "Offline"
+
+
+[service_webserver]
+# Check if a webserver is running on port 80. You can replace the ip with a domain name.
+# You can check different services changing the port number. Some examples:
+# 80 HTTP Webserver
+# 443 HTTPS Webserver
+# 21 FTP
+# 22 SSH
+# 5900 VNC (Linux remote desktop)
+# 3389 RDP (Windows remote desktop)
+DISABLED=True
+ALARM_STRING_NOT_EQUAL=Online
+COMMAND=nc -z -w 3 192.168.1.123 80 > /dev/null && echo "Online" || echo "Offline"
+
+
+[dummy_always_alarm]
+# A dummy check that is always in alarm. Useful for testing notifications.
+DISABLED=True
+ALARM_STRING_EQUAL=Core meltdown!
+COMMAND=echo "Core meltdown!"

healthcheck/healthcheck.py

@@ -46,13 +46,18 @@ import socket
 import getpass
 import re
 import locale
+import json
 
 
 NAME = 'healthcheck'
 VERSION = '0.1'
 DESCRIPTION = 'A simple server monitoring software'
-EMAIL_SUBJECT_TPL = 'Host {} failed health check for {}'
+EMAIL_START_SUBJECT_TPL = '\U0001F6A8 {}: {} ALARM!'
-EMAIL_MESSAGE_TPL = 'Alarm for sensor {} on host {} on {}: {}'
+EMAIL_START_MESSAGE_TPL = 'Alarm for sensor {} on host {} on {}: {}'
+EMAIL_END_SUBJECT_TPL = '\u2705 {}: {} OK'
+EMAIL_END_MESSAGE_TPL = 'Alarm ceased for sensor {} on host {} on {}'
+# Healthcheck saves the current status (alarms triggered, last run... in this file)
+STATUS_FILE = '/tmp/healthcheck.tmp'
 
 class Main:
 
@@ -78,6 +83,10 @@ class Main:
 	def run(self, dryRun):
 		''' Runs the health checks '''
 
+		# Load status
+		status = Status()
+
+		# Run checks based o the config
 		for section in self.config:
 			if section == 'DEFAULT':
 				continue
@@ -85,6 +94,7 @@ class Main:
 			s = Settings(section, self.config)
 			if s.disabled:
 				self._log.info('Ignoring disabled check "{}"'.format(section))
+				status.unsetAlarm(section)
 				continue
 
 			self._log.info('Checking "{}"'.format(section))
@@ -93,11 +103,37 @@ class Main:
 			if error:
 				# Alarm!
 				logging.warning('Alarm for {}: {}!'.format(section, error))
-				if not dryRun:
+				if self.shouldNotify(section, s, status):
-					if s.mailto:
+					status.setAlarm(section)
-						self.sendMail(s, error)
+					if not dryRun:
-					if s.alarmCommand:
+						if s.mailto:
-						self.executeAlarmCommand(s, error)
+							self.sendAlmStartMail(s, error)
+						if s.alarmCommand:
+							self.executeAlarmCommand(s, error)
+			elif status.getAlarmTriggeredTimestamp(section) is not None:
+				logging.info('Alarm ceased for {}: OK!'.format(section))
+				if s.notify_alarm_end:
+					self.sendAlmEndMail(s)
+				status.unsetAlarm(section)
+
+		# Save updated status
+		status.save()
+
+	def shouldNotify(self, section, settings, status):
+		almTriggeredTime = status.getAlarmTriggeredTimestamp(section)
+		# Notify if alarm just started
+		if almTriggeredTime is None:
+			return True
+
+		# Notify if NOTIFY=EVERY_RUN
+		if settings.notify == 'EVERY_RUN':
+			return True
+
+		# Notify if time elapsed
+		if settings.notify == 'ONCE_IN_MINUTES' and (time.time() - almTriggeredTime) > (settings.notify_minutes * 60):
+			return True
+
+		return False
 
 	# Calls the provided command, checks the value parsing it with the provided regexp
 	# and returns an error string, or null if the value is within its limits
@@ -112,12 +148,15 @@ class Main:
 		stdout = ""
 		ret = subprocess.run(config.command, stdout=subprocess.PIPE, stderr=subprocess.PIPE, shell=True)
 		if ret.stderr:
-			self._log.info('{} subprocess stderr:\n{}', config.command, ret.stderr.decode())
+			self._log.info('{} subprocess stderr:\n{}'.format(config.command, ret.stderr.decode()))
 		if ret.stdout:
 			stdout = ret.stdout.decode()
-			self._log.debug('{} subprocess stdout:\n{}', config.command, stdout)
+			self._log.debug('{} subprocess stdout:\n{}'.format(config.command, stdout))
 		if ret.returncode != 0:
-			return 'subprocess {} exited with error code {}'.format(config.command, ret.returncode)
+			return 'the command exited with error code {} {}'.format(
+				ret.returncode,
+				'and error message "{}"'.format(ret.stderr.decode().strip()) if ret.stderr else ''
+			)
 		
 		# Parse result with regex
 		match = re.search(config.regexp, stdout, re.MULTILINE)
@@ -143,15 +182,34 @@ class Main:
 		if config.alarm_value_less_than and locale.atof(detectedValue) < float(config.alarm_value_less_than):
 			return 'value is {}, but should be greater than {}'.format(locale.atof(detectedValue), config.alarm_value_less_than)
 
-	def sendMail(self, s, error):
+	def sendAlmStartMail(self, s, error):
+		subject = EMAIL_START_SUBJECT_TPL.format(self.hostname, s.name)
+		body = EMAIL_START_MESSAGE_TPL.format(
+			s.name,
+			self.hostname,
+			time.strftime("%a, %d %b %Y %H:%M:%S"),
+			error
+		)
+		self.sendMail(s, subject, body)
+
+	def sendAlmEndMail(self, s):
+		subject = EMAIL_END_SUBJECT_TPL.format(self.hostname, s.name)
+		body = EMAIL_END_MESSAGE_TPL.format(
+			s.name,
+			self.hostname,
+			time.strftime("%a, %d %b %Y %H:%M:%S")
+		)
+		self.sendMail(s, subject, body)
+
+	def sendMail(self, s, subject, body):
 		if s.smtphost:
-			logging.info("Sending alarm email to %s via %s", s.mailto, s.smtphost)
+			logging.info("Sending email to %s via %s", s.mailto, s.smtphost)
 		else:
-			logging.info("Sending alarm email to %s using local smtp", s.mailto)
+			logging.info("Sending email to %s using local smtp", s.mailto)
 
 		# Create main message
 		msg = MIMEMultipart()
-		msg['Subject'] = EMAIL_SUBJECT_TPL.format(self.hostname, s.name)
+		msg['Subject'] = subject
 		if s.mailfrom:
 			m_from = s.mailfrom
 		else:
@@ -161,12 +219,6 @@ class Main:
 		msg.preamble = 'This is a multi-part message in MIME format.'
 
 		# Add base text
-		body = EMAIL_MESSAGE_TPL.format(
-			s.name,
-			self.hostname,
-			time.strftime("%a, %d %b %Y %H:%M:%S"),
-			error
-		)
 		txt = MIMEText(body)
 		msg.attach(txt)
 
@@ -204,6 +256,34 @@ class Main:
 			self._log.error('subprocess {} exited with error code {}'.format(cmdToRun, ret.returncode))
 
 
+class Status:
+	''' Represents the current status (alarms triggered, last run...) '''
+
+	def __init__(self):
+		try:
+			with open(STATUS_FILE, 'r') as openfile:
+				self.status = json.load(openfile)
+		except FileNotFoundError:
+			self.status =  {
+				'lastRun': 0,	# unix time in seconds
+				'alarms': {},	# key-value, alarmName : alarmTriggeredTimestamp
+			}
+
+	def save(self):
+		self.status['lastRun'] = time.time()
+		jo = json.dumps(self.status)
+		with open(STATUS_FILE, "w") as outfile:
+			outfile.write(jo)
+
+	def setAlarm(self, almName):
+		self.status['alarms'][almName] = time.time()
+
+	def unsetAlarm(self, almName):
+		self.status['alarms'].pop(almName, None)
+
+	def getAlarmTriggeredTimestamp(self, almName):
+		return self.status['alarms'].get(almName, None)
+
 
 class Settings:
 	''' Represents settings for a check '''
@@ -241,6 +321,10 @@ class Settings:
 		self.alarm_value_not_equal = self.getStr(name, 'ALARM_VALUE_NOT_EQUAL', None)
 		self.alarm_value_more_than = self.getStr(name, 'ALARM_VALUE_MORE_THAN', None)
 		self.alarm_value_less_than = self.getStr(name, 'ALARM_VALUE_LESS_THAN', None)
+		## Notification policy
+		self.notify = self.getEnum(name, 'NOTIFY', 'EVERY_RUN', ['EVERY_RUN', 'START', 'ONCE_IN_MINUTES'])
+		self.notify_minutes = self.getInt(name, 'NOTIFY_MINUTES', 0)
+		self.notify_alarm_end = self.getBoolean(name, 'NOTIFY_ALARM_END', True)
 		## Command to obtain the value for comparation
 		self.command = self.getStr(name, 'COMMAND', None)
 		## Regexp to extract value from command output (default to match full string)
@@ -252,12 +336,20 @@ class Settings:
 		except configparser.NoOptionError:
 			return defaultValue
 
+	def getInt(self, name, key, defaultValue):
+		return int(self.getStr(name, key, defaultValue))
+
 	def getBoolean(self, name, key, defaultValue):
 		try:
 			return self.config.getboolean(name, key)
 		except configparser.NoOptionError:
 			return defaultValue
 
+	def getEnum(self, name, key, defaultValue, values):
+		val = self.getStr(name, key, defaultValue)
+		if not val in values:
+			raise ValueError("Invalid value {} for configuration {}: expected one of {}".format(val, key, ', '.join(values)))
+		return val
 
 
 if __name__ == '__main__':