Run all checks with a signal (#44)

* run all checks with signal

Sending a SIGURG triggers an immediate, additional run of all checks.

README.rst

@@ -608,6 +608,8 @@ You can launch it by supplying a configuration file and a directory with configu
 
 At the root of the project there is System V init and a Systemd unit file for proper integration with OS startup tools.
 
+Sending a ``SIGURG`` signal to a running anycast-healthchecker process will trigger an immediate, additional (not changing the regular interval) execution of all active checks. For services with ``check_rise`` and/or ``check_fail`` set to ``1``, this can be used to make external events faster advertise and/or withdraw their prefixes.
+
 Systemd and SysVinit integration
 ################################
 

anycast_healthchecker/healthchecker.py

@@ -6,6 +6,7 @@
 import sys
 import logging
 from queue import Queue
+import threading
 
 from anycast_healthchecker import PROGRAM_NAME, METRIC_PREFIX
 from anycast_healthchecker.servicecheck import ServiceCheck
@@ -76,6 +77,8 @@ def __init__(self, config, bird_configuration):
             )
             self.ip_prefixes[ip_version] = _ip_prefixes
 
+        self._urgent_event = threading.Event()
+
         self.log.info('initialize healthchecker')
 
     def _update_bird_conf_file(self, operation):
@@ -247,7 +250,8 @@ def run(self):
                                        metric_state,
                                        metric_check_duration,
                                        metric_check_ip_assignment,
-                                       metric_check_timeout)
+                                       metric_check_timeout,
+                                       self._urgent_event)
                 _thread.start()
 
         # Stay running until we are stopped
@@ -277,3 +281,9 @@ def run(self):
                         self.bird_configuration[ip_version]['reconfigure_cmd'])
                 else:
                     run_custom_bird_reconfigure(operation)
+
+    def run_all_checks_now(self):
+        """Immediately run all checks. This does not change the usual interval."""
+        self.log.info("Immediatly running all checks")
+        self._urgent_event.set()
+        self._urgent_event.clear()

anycast_healthchecker/main.py

@@ -94,6 +94,11 @@ def main():
 
     # Create our master process.
     checker = healthchecker.HealthChecker(config, bird_configuration)
+
+    # Register our SIGURG handler to immediately trigger all checks.
+    signal.signal(signal.SIGURG, lambda signum, frame: checker.run_all_checks_now())
+
+    # and start working
     logger.info("starting %s version %s", PROGRAM_NAME, __version__)
     checker.run()
 

anycast_healthchecker/servicecheck.py

@@ -37,7 +37,7 @@ class ServiceCheck(Thread):
 
     def __init__(self, service, config, action, splay_startup, metric_state,
                  metric_check_duration, metric_check_ip_assignment,
-                 metric_check_timeout):
+                 metric_check_timeout, urgent_event):
         """Set the name of thread to be the name of the service."""
         super(ServiceCheck, self).__init__()
         self.name = service  # Used by Thread()
@@ -85,6 +85,7 @@ def __init__(self, service, config, action, splay_startup, metric_state,
             bird_reconfigure_cmd=config.get('custom_bird_reconfigure_cmd',
                                             None)
         )
+        self.urgent_event = urgent_event
         self.log.info("loading check for %s", self.name, extra=self.extra)
 
         self.metric_state = metric_state
@@ -365,4 +366,4 @@ def _run(self):
             if sleep < 0:
                 sleep += interval
             self.log.debug("sleeping for %.3fsecs", sleep, extra=self.extra)
-            time.sleep(sleep)
+            self.urgent_event.wait(timeout=sleep)