feat(autofix): Duplicate check for automation runs (#102933)

## PR Details
+ For the V0 launch of triage signals we are changing the conditions for
starting an autofix run.
+ More details explanation here:
https://www.notion.so/sentry/Triage-Signals-V0-Technical-Implementation-Details-2a18b10e4b5d8086a7ceddaf4194849a?source=copy_link#2a18b10e4b5d80d2ab2ce33d12db52e0
+ As mentioned in the doc we can't rely on seer_fixability_score alone
going forward and hence this PR adds cache based check +
seer_autofix_last_triggered as a check.
+ A new column is overkill for this. @wedamija was right to point it out
[in this PR.](#102898) Adding a
redis cache based check is enough.

Author: Mihir-Mavalankar

src/sentry/tasks/post_process.py

@@ -1628,7 +1628,7 @@ def kick_off_seer_automation(job: PostProcessJob) -> None:
     event = job["event"]
     group = event.group
 
-    # Only run on issues with no existing scan
+    # Only run on issues with no existing scan - TODO: Update condition for triage signals V0
     if group.seer_fixability_score is not None:
         return
 
@@ -1660,6 +1660,13 @@ def kick_off_seer_automation(job: PostProcessJob) -> None:
     ):
         return
 
+    # Check if automation has already been queued or completed for this group
+    # seer_autofix_last_triggered is set when trigger_autofix is successfully started.
+    # Use cache with short TTL to hold lock for a short since it takes a few minutes to set seer_autofix_last_triggeredes
+    cache_key = f"seer_automation_queued:{group.id}"
+    if cache.get(cache_key) or group.seer_autofix_last_triggered is not None:
+        return
+
     # Don't run if there's already a task in progress for this issue
     lock_key, lock_name = get_issue_summary_lock_key(group.id)
     lock = locks.get(lock_key, duration=1, name=lock_name)
@@ -1684,6 +1691,11 @@ def kick_off_seer_automation(job: PostProcessJob) -> None:
     if is_seer_scanner_rate_limited(project, group.organization):
         return
 
+    # cache.add uses Redis SETNX which atomically sets the key only if it doesn't exist
+    # Returns False if another process already set the key, ensuring only one process proceeds
+    if not cache.add(cache_key, True, timeout=600):  # 10 minute
+        return
+
     start_seer_automation.delay(group.id)
 
 

tests/sentry/tasks/test_post_process.py

@@ -3053,6 +3053,137 @@ def test_kick_off_seer_automation_with_hide_ai_features_enabled(
 
         mock_start_seer_automation.assert_not_called()
 
+    @patch(
+        "sentry.seer.seer_setup.get_seer_org_acknowledgement",
+        return_value=True,
+    )
+    @patch("sentry.tasks.autofix.start_seer_automation.delay")
+    @with_feature("organizations:gen-ai-features")
+    def test_kick_off_seer_automation_skips_when_seer_autofix_last_triggered_set(
+        self, mock_start_seer_automation, mock_get_seer_org_acknowledgement
+    ):
+        """Test that automation is skipped when group.seer_autofix_last_triggered is already set"""
+        self.project.update_option("sentry:seer_scanner_automation", True)
+        event = self.create_event(
+            data={"message": "testing"},
+            project_id=self.project.id,
+        )
+
+        # Set seer_autofix_last_triggered on the group to simulate autofix already triggered
+        group = event.group
+        group.seer_autofix_last_triggered = timezone.now()
+        group.save()
+
+        self.call_post_process_group(
+            is_new=True,
+            is_regression=False,
+            is_new_group_environment=True,
+            event=event,
+        )
+
+        mock_start_seer_automation.assert_not_called()
+
+    @patch(
+        "sentry.seer.seer_setup.get_seer_org_acknowledgement",
+        return_value=True,
+    )
+    @patch("sentry.tasks.autofix.start_seer_automation.delay")
+    @with_feature("organizations:gen-ai-features")
+    def test_kick_off_seer_automation_skips_when_cache_key_exists(
+        self, mock_start_seer_automation, mock_get_seer_org_acknowledgement
+    ):
+        """Test that automation is skipped when cache key indicates it's already queued"""
+        self.project.update_option("sentry:seer_scanner_automation", True)
+        event = self.create_event(
+            data={"message": "testing"},
+            project_id=self.project.id,
+        )
+
+        # Set cache key to simulate automation already queued
+        cache_key = f"seer_automation_queued:{event.group.id}"
+        cache.set(cache_key, True, timeout=600)
+
+        self.call_post_process_group(
+            is_new=True,
+            is_regression=False,
+            is_new_group_environment=True,
+            event=event,
+        )
+
+        mock_start_seer_automation.assert_not_called()
+
+        # Cleanup
+        cache.delete(cache_key)
+
+    @patch(
+        "sentry.seer.seer_setup.get_seer_org_acknowledgement",
+        return_value=True,
+    )
+    @patch("sentry.tasks.autofix.start_seer_automation.delay")
+    @with_feature("organizations:gen-ai-features")
+    def test_kick_off_seer_automation_uses_atomic_cache_add(
+        self, mock_start_seer_automation, mock_get_seer_org_acknowledgement
+    ):
+        """Test that cache.add atomic operation prevents race conditions"""
+        self.project.update_option("sentry:seer_scanner_automation", True)
+        event = self.create_event(
+            data={"message": "testing"},
+            project_id=self.project.id,
+        )
+
+        cache_key = f"seer_automation_queued:{event.group.id}"
+
+        with patch("sentry.tasks.post_process.cache") as mock_cache:
+            # Simulate cache.get returning None (not in cache)
+            # but cache.add returning False (another process set it)
+            mock_cache.get.return_value = None
+            mock_cache.add.return_value = False
+
+            self.call_post_process_group(
+                is_new=True,
+                is_regression=False,
+                is_new_group_environment=True,
+                event=event,
+            )
+
+            # Should check cache but not call automation due to cache.add returning False
+            mock_cache.get.assert_called()
+            mock_cache.add.assert_called_once_with(cache_key, True, timeout=600)
+            mock_start_seer_automation.assert_not_called()
+
+    @patch(
+        "sentry.seer.seer_setup.get_seer_org_acknowledgement",
+        return_value=True,
+    )
+    @patch("sentry.tasks.autofix.start_seer_automation.delay")
+    @with_feature("organizations:gen-ai-features")
+    def test_kick_off_seer_automation_proceeds_when_cache_add_succeeds(
+        self, mock_start_seer_automation, mock_get_seer_org_acknowledgement
+    ):
+        """Test that automation proceeds when cache.add succeeds (no race condition)"""
+        self.project.update_option("sentry:seer_scanner_automation", True)
+        event = self.create_event(
+            data={"message": "testing"},
+            project_id=self.project.id,
+        )
+
+        # Ensure seer_autofix_last_triggered is not set
+        assert event.group.seer_autofix_last_triggered is None
+
+        self.call_post_process_group(
+            is_new=True,
+            is_regression=False,
+            is_new_group_environment=True,
+            event=event,
+        )
+
+        # Should successfully queue automation
+        mock_start_seer_automation.assert_called_once_with(event.group.id)
+
+        # Cleanup
+        cache_key = f"seer_automation_queued:{event.group.id}"
+        cache.delete(cache_key)
+
 
 class PostProcessGroupErrorTest(
     TestCase,