Fix errant changelog entries when executing a script without committing

netbox/extras/middleware.py

@@ -9,11 +9,12 @@ from django.utils import timezone
 from django.utils.functional import curry
 from django_prometheus.models import model_deletes, model_inserts, model_updates
 
-from extras.webhooks import enqueue_webhooks
 from .constants import (
     OBJECTCHANGE_ACTION_CREATE, OBJECTCHANGE_ACTION_DELETE, OBJECTCHANGE_ACTION_UPDATE,
 )
 from .models import ObjectChange
+from .signals import purge_changelog
+from .webhooks import enqueue_webhooks
 
 _thread_locals = threading.local()
 
@@ -30,6 +31,10 @@ def cache_changed_object(instance, **kwargs):
 
 def _record_object_deleted(request, instance, **kwargs):
 
+    # TODO: Can we cache deletions for later processing like we do for saves? Currently this will trigger an exception
+    # when trying to serialize ManyToMany relations after the object has been deleted. This should be doable if we alter
+    # log_change() to return ObjectChanges to be saved rather than saving them directly.
+
     # Record that the object was deleted
     if hasattr(instance, 'log_change'):
         instance.log_change(request.user, request.id, OBJECTCHANGE_ACTION_DELETE)
@@ -41,6 +46,13 @@ def _record_object_deleted(request, instance, **kwargs):
     model_deletes.labels(instance._meta.model_name).inc()
 
 
+def purge_objectchange_cache(sender, **kwargs):
+    """
+    Delete any queued object changes waiting to be written.
+    """
+    _thread_locals.changed_objects = None
+
+
 class ObjectChangeMiddleware(object):
     """
     This middleware performs three functions in response to an object being created, updated, or deleted:
@@ -74,9 +86,21 @@ class ObjectChangeMiddleware(object):
         post_save.connect(cache_changed_object, dispatch_uid='record_object_saved')
         post_delete.connect(record_object_deleted, dispatch_uid='record_object_deleted')
 
+        # Provide a hook for purging the change cache
+        purge_changelog.connect(purge_objectchange_cache)
+
         # Process the request
         response = self.get_response(request)
 
+        # If the change cache has been purged (e.g. due to an exception) abort the logging of all changes resulting from
+        # this request.
+        if _thread_locals.changed_objects is None:
+
+            # Delete ObjectChanges representing deletions, since these have already been written
+            ObjectChange.objects.filter(request_id=request.id).delete()
+
+            return response
+
         # Create records for any cached objects that were created/updated.
         for obj, action in _thread_locals.changed_objects:
 

netbox/extras/scripts.py

@@ -18,6 +18,7 @@ from ipam.formfields import IPFormField
 from utilities.exceptions import AbortTransaction
 from .constants import LOG_DEFAULT, LOG_FAILURE, LOG_INFO, LOG_SUCCESS, LOG_WARNING
 from .forms import ScriptForm
+from .signals import purge_changelog
 
 
 __all__ = [
@@ -310,6 +311,8 @@ def run_script(script, data, files, commit=True):
         commit = False
     finally:
         if not commit:
+            # Delete all pending changelog entries
+            purge_changelog.send(Script)
             script.log_info(
                 "Database changes have been reverted automatically."
             )

netbox/extras/signals.py

@@ -1,7 +1,12 @@
 from cacheops.signals import cache_invalidated, cache_read
+from django.dispatch import Signal
 from prometheus_client import Counter
 
 
+#
+# Caching
+#
+
 cacheops_cache_hit = Counter('cacheops_cache_hit', 'Number of cache hits')
 cacheops_cache_miss = Counter('cacheops_cache_miss', 'Number of cache misses')
 cacheops_cache_invalidated = Counter('cacheops_cache_invalidated', 'Number of cache invalidations')
@@ -20,3 +25,10 @@ def cache_invalidated_collector(sender, obj_dict, **kwargs):
 
 cache_read.connect(cache_read_collector)
 cache_invalidated.connect(cache_invalidated_collector)
+
+
+#
+# Change logging
+#
+
+purge_changelog = Signal()