[refactor] Improve efficiency of check_missing_pricing (#12572)

src/backend/InvenTree/part/tasks.py

@@ -17,6 +17,7 @@ import InvenTree.helpers_model
 from common.settings import get_global_setting
 from InvenTree.tasks import (
     ScheduledTask,
+    batch_offload_tasks,
     check_daily_holdoff,
     offload_task,
     record_task_success,
@@ -265,45 +266,51 @@ def check_missing_pricing(limit=250):
         # Task does not run if the interval is zero
         return
 
-    # Find parts for which pricing information has never been updated
+    # Scheduling each part's pricing update calls offload_task(), which (outside of a batch)
-    results = PartPricing.objects.filter(updated=None)[:limit]
+    # checks the entire task queue for duplicates on every call. Batching collapses all of
+    # this run's scheduling into a single bulk write instead, avoiding that per-call scan.
+    with batch_offload_tasks():
+        # Find parts for which pricing information has never been updated
+        results = PartPricing.objects.filter(updated=None)[:limit]
 
-    if results.count() > 0:
+        if results.count() > 0:
-        logger.info('Found %s parts with empty pricing', results.count())
+            logger.info('Found %s parts with empty pricing', results.count())
 
-        for pp in results:
+            for pp in results:
-            pp.schedule_for_update()
+                pp.schedule_for_update()
 
-    stale_date = datetime.now().date() - timedelta(days=days)
+        stale_date = datetime.now().date() - timedelta(days=days)
 
-    results = PartPricing.objects.filter(updated__lte=stale_date)[:limit]
+        results = PartPricing.objects.filter(updated__lte=stale_date)[:limit]
 
-    if results.count() > 0:
+        if results.count() > 0:
-        logger.info('Found %s stale pricing entries', results.count())
+            logger.info('Found %s stale pricing entries', results.count())
 
-        for pp in results:
+            for pp in results:
-            pp.schedule_for_update()
+                pp.schedule_for_update()
 
-    # Find any pricing data which is in the wrong currency
+        # Find any pricing data which is in the wrong currency
-    currency = common.currency.currency_code_default()
+        currency = common.currency.currency_code_default()
-    results = PartPricing.objects.exclude(currency=currency)
+        results = PartPricing.objects.exclude(currency=currency)[:limit]
 
-    if results.count() > 0:
+        if results.count() > 0:
-        logger.info('Found %s pricing entries in the wrong currency', results.count())
+            logger.info(
+                'Found %s pricing entries in the wrong currency', results.count()
+            )
 
-        for pp in results:
+            for pp in results:
-            pp.schedule_for_update()
+                pp.schedule_for_update()
 
-    # Find any parts which do not have pricing information
+        # Find any parts which do not have pricing information
-    results = Part.objects.filter(pricing_data=None)[:limit]
+        results = Part.objects.filter(pricing_data=None)[:limit]
 
-    if results.count() > 0:
+        if results.count() > 0:
-        logger.info('Found %s parts without pricing', results.count())
+            logger.info('Found %s parts without pricing', results.count())
 
-        for p in results:
+            for p in results:
-            pricing = p.pricing
+                pricing = p.pricing
-            pricing.save()
+                pricing.save()
-            pricing.schedule_for_update()
+                pricing.schedule_for_update()
 
 
 @tracer.start_as_current_span('scheduled_stocktake_reports')

src/backend/InvenTree/part/test_pricing.py

@@ -1,5 +1,7 @@
 """Unit tests for Part pricing calculations."""
 
+from unittest import mock
+
 from django.core.exceptions import ObjectDoesNotExist
 from django.test.utils import override_settings
 
@@ -431,6 +433,38 @@ class PartPricingTests(InvenTreeTestCase):
         # Check that PartPricing objects have been created
         self.assertEqual(part.models.PartPricing.objects.count(), 101)
 
+    def test_check_missing_pricing_batches_scheduling(self):
+        """check_missing_pricing() must batch its scheduling calls, not scan the task queue per part.
+
+        Regression test: without batching, each PartPricing.schedule_for_update() call triggers
+        offload_task() -> check_existing_task(), which scans and unpickles every row in the task
+        queue to look for a duplicate. Calling this once per part inside check_missing_pricing's
+        loops made scheduling cost grow with the queue backlog, which is what caused a background
+        worker timeout in production (Sentry: "Task exceeded maximum timeout value (90 seconds)"
+        raised from check_existing_task).
+        """
+        from part.tasks import check_missing_pricing
+
+        # Create some parts (deliberately not using TESTING_PRICING here, so that
+        # schedule_for_update() actually offloads via the task queue rather than running inline)
+        for ii in range(20):
+            part.models.Part.objects.create(
+                name=f'Part_{ii}', description='A test part'
+            )
+
+        # Ensure there is no pricing data
+        part.models.PartPricing.objects.all().delete()
+
+        with self.captureOnCommitCallbacks(execute=True):
+            with mock.patch('InvenTree.tasks.check_existing_task') as mock_check:
+                check_missing_pricing()
+
+        # Scheduling was batched - the per-call duplicate-check scan never ran
+        mock_check.assert_not_called()
+
+        # PartPricing objects were still created and scheduled for a background update
+        self.assertEqual(part.models.PartPricing.objects.count(), 21)
+
     @override_settings(TESTING_PRICING=True)
     def test_delete_part_with_stock_items(self):
         """Test deleting a part instance with stock items.