Merge pull request #2836 from ryanpetrello/better-dispatch-retries

add additional DB retry logic to the callback receiver

Reviewed-by: https://github.com/softwarefactory-project-zuul[bot]

awx/main/dispatch/worker/base.py

@@ -7,6 +7,7 @@ import signal
 from uuid import UUID
 from Queue import Empty as QueueEmpty
 
+from django import db
 from kombu import Producer
 from kombu.mixins import ConsumerMixin
 
@@ -128,6 +129,10 @@ class BaseWorker(object):
                 logger.error("Exception on worker {}, restarting: ".format(idx) + str(e))
                 continue
             try:
+                for conn in db.connections.all():
+                    # If the database connection has a hiccup during the prior message, close it
+                    # so we can establish a new connection
+                    conn.close_if_unusable_or_obsolete()
                 self.perform_work(body, *args)
             finally:
                 if 'uuid' in body:

awx/main/dispatch/worker/callback.py

@@ -1,7 +1,5 @@
 import logging
 import time
-import os
-import signal
 import traceback
 
 from django.conf import settings
@@ -110,8 +108,7 @@ class CallbackBrokerWorker(BaseWorker):
                     break
                 except (OperationalError, InterfaceError, InternalError):
                     if retries >= self.MAX_RETRIES:
-                        logger.exception('Worker could not re-establish database connectivity, shutting down gracefully: Job {}'.format(job_identifier))
-                        os.kill(os.getppid(), signal.SIGINT)
+                        logger.exception('Worker could not re-establish database connectivity, giving up on event for Job {}'.format(job_identifier))
                         return
                     delay = 60 * retries
                     logger.exception('Database Error Saving Job Event, retry #{i} in {delay} seconds:'.format(

awx/main/dispatch/worker/task.py

@@ -5,7 +5,6 @@ import sys
 import traceback
 
 import six
-from django import db
 
 from awx.main.tasks import dispatch_startup, inform_cluster_of_shutdown
 
@@ -75,10 +74,6 @@ class TaskWorker(BaseWorker):
             'task': u'awx.main.tasks.RunProjectUpdate'
         }
         '''
-        for conn in db.connections.all():
-            # If the database connection has a hiccup during at task, close it
-            # so we can establish a new connection
-            conn.close_if_unusable_or_obsolete()
         result = None
         try:
             result = self.run_callable(body)