Merge pull request #6093 from ryanpetrello/remove-beat-thread

switch the periodic scheduler to a background process (instead of a thread) to avoid a cpython bug

Reviewed-by: https://github.com/apps/softwarefactory-project-zuul

awx/main/dispatch/periodic.py

@@ -1,6 +1,7 @@
 import logging
-import threading
+import os
 import time
+from multiprocessing import Process
 
 from django.conf import settings
 from django.db import connections
@@ -14,33 +15,36 @@ logger = logging.getLogger('awx.main.dispatch.periodic')
 class Scheduler(Scheduler):
 
     def run_continuously(self):
-        cease_continuous_run = threading.Event()
         idle_seconds = max(
             1,
             min(self.jobs).period.total_seconds() / 2
         )
 
-        class ScheduleThread(threading.Thread):
+        def run():
-            @classmethod
+            ppid = os.getppid()
-            def run(cls):
+            logger.warn(f'periodic beat started')
-                while not cease_continuous_run.is_set():
+            while True:
-                    try:
+                if os.getppid() != ppid:
-                        for conn in connections.all():
+                    # if the parent PID changes, this process has been orphaned
-                            # If the database connection has a hiccup, re-establish a new
+                    # via e.g., segfault or sigkill, we should exit too
-                            # connection
+                    pid = os.getpid()
-                            conn.close_if_unusable_or_obsolete()
+                    logger.warn(f'periodic beat exiting gracefully pid:{pid}')
-                        self.run_pending()
+                    raise SystemExit()
-                    except Exception:
+                try:
-                        logger.exception(
+                    for conn in connections.all():
-                            'encountered an error while scheduling periodic tasks'
+                        # If the database connection has a hiccup, re-establish a new
-                        )
+                        # connection
-                    time.sleep(idle_seconds)
+                        conn.close_if_unusable_or_obsolete()
-                logger.debug('periodic thread exiting...')
+                    self.run_pending()
+                except Exception:
+                    logger.exception(
+                        'encountered an error while scheduling periodic tasks'
+                    )
+                time.sleep(idle_seconds)
 
-        thread = ScheduleThread()
+        process = Process(target=run)
-        thread.daemon = True
+        process.daemon = True
-        thread.start()
+        process.start()
-        return cease_continuous_run
 
 
 def run_continuously():
@@ -49,4 +53,4 @@ def run_continuously():
         apply_async = TaskWorker.resolve_callable(task['task']).apply_async
         total_seconds = task['schedule'].total_seconds()
         scheduler.every(total_seconds).seconds.do(apply_async)
-    return scheduler.run_continuously()
+    scheduler.run_continuously()

awx/main/management/commands/run_dispatcher.py

@@ -53,7 +53,7 @@ class Command(BaseCommand):
 
         # spawn a daemon thread to periodically enqueues scheduled tasks
         # (like the node heartbeat)
-        cease_continuous_run = periodic.run_continuously()
+        periodic.run_continuously()
 
         reaper.reap()
         consumer = None
@@ -87,7 +87,6 @@ class Command(BaseCommand):
                 )
                 consumer.run()
             except KeyboardInterrupt:
-                cease_continuous_run.set()
                 logger.debug('Terminating Task Dispatcher')
                 if consumer:
                     consumer.stop()