fix: fixing exclude region inclusion (#43602)

* fix: fixing exclude region issues

* updated CHANGELOG.md

* fix: fixing bug in the implementation

* fix: refactoring

* fix: fixing tests

* fix: fixing pylint

* fix: fixing pylint

* fix: fixing pylint

* fix: fixing pylint

* fix: fixing logic

* fix: cleaning up code

Author: dibahlfi

sdk/cosmos/azure-cosmos/CHANGELOG.md

@@ -7,8 +7,10 @@
 #### Breaking Changes
 
 #### Bugs Fixed
+* Fixed bug where customer provided excluded region was not always being honored during certain transient failures. See [PR 43602](https://github.com/Azure/azure-sdk-for-python/pull/43602)
 
 #### Other Changes
+* Enhanced logging to ensure when a region is marked unavailable we have the proper context. See [PR 43602](https://github.com/Azure/azure-sdk-for-python/pull/43602)
 
 ### 4.14.0 (2025-10-13)
 This version and all future versions will require Python 3.9+.

sdk/cosmos/azure-cosmos/azure/cosmos/_endpoint_discovery_retry_policy.py

@@ -71,15 +71,16 @@ def ShouldRetry(self, exception):  # pylint: disable=unused-argument
         self.failover_retry_count += 1
 
         if self.request.location_endpoint_to_route:
+            context = self.__class__.__name__
             if _OperationType.IsReadOnlyOperation(self.request.operation_type):
                 # Mark current read endpoint as unavailable
                 self.global_endpoint_manager.mark_endpoint_unavailable_for_read(
                     self.request.location_endpoint_to_route,
-                    True)
+                    True, context)
             else:
                 self.global_endpoint_manager.mark_endpoint_unavailable_for_write(
                     self.request.location_endpoint_to_route,
-                    True)
+                    True, context)
 
         # set the refresh_needed flag to ensure that endpoint list is
         # refreshed with new writable and readable locations

sdk/cosmos/azure-cosmos/azure/cosmos/_global_endpoint_manager.py

@@ -74,11 +74,11 @@ def _resolve_service_endpoint(
     ) -> str:
         return self.location_cache.resolve_service_endpoint(request)
 
-    def mark_endpoint_unavailable_for_read(self, endpoint, refresh_cache):
-        self.location_cache.mark_endpoint_unavailable_for_read(endpoint, refresh_cache)
+    def mark_endpoint_unavailable_for_read(self, endpoint, refresh_cache, context: str):
+        self.location_cache.mark_endpoint_unavailable_for_read(endpoint, refresh_cache, context)
 
-    def mark_endpoint_unavailable_for_write(self, endpoint, refresh_cache):
-        self.location_cache.mark_endpoint_unavailable_for_write(endpoint, refresh_cache)
+    def mark_endpoint_unavailable_for_write(self, endpoint, refresh_cache, context: str):
+        self.location_cache.mark_endpoint_unavailable_for_write(endpoint, refresh_cache, context)
 
     def get_ordered_write_locations(self):
         return self.location_cache.get_ordered_write_locations()
@@ -96,14 +96,15 @@ def force_refresh_on_startup(self, database_account):
     def update_location_cache(self):
         self.location_cache.update_location_cache()
 
-    def _mark_endpoint_unavailable(self, endpoint: str):
+    def _mark_endpoint_unavailable(self, endpoint: str, context: str):
         """Marks an endpoint as unavailable for the appropriate operations.
         :param str endpoint: The endpoint to mark as unavailable.
+        :param str context: The context for marking the endpoint as unavailable.
         """
         write_endpoints = self.location_cache.get_all_write_endpoints()
-        self.mark_endpoint_unavailable_for_read(endpoint, False)
+        self.mark_endpoint_unavailable_for_read(endpoint, False, context)
         if endpoint in write_endpoints:
-            self.mark_endpoint_unavailable_for_write(endpoint, False)
+            self.mark_endpoint_unavailable_for_write(endpoint, False, context)
 
     def refresh_endpoint_list(self, database_account, **kwargs):
         if current_time_millis() - self.last_refresh_time > self.refresh_time_interval_in_ms:
@@ -159,7 +160,7 @@ def _GetDatabaseAccount(self, **kwargs) -> Tuple[DatabaseAccount, str]:
                     self.location_cache.mark_endpoint_available(locational_endpoint)
                     return database_account, locational_endpoint
                 except (exceptions.CosmosHttpResponseError, AzureError):
-                    self._mark_endpoint_unavailable(locational_endpoint)
+                    self._mark_endpoint_unavailable(locational_endpoint, "_GetDatabaseAccount")
             raise
 
     def _endpoints_health_check(self, **kwargs):
@@ -194,7 +195,7 @@ def _endpoints_health_check(self, **kwargs):
                     success_count += 1
                     self.location_cache.mark_endpoint_available(endpoint)
                 except (exceptions.CosmosHttpResponseError, AzureError):
-                    self._mark_endpoint_unavailable(endpoint)
+                    self._mark_endpoint_unavailable(endpoint, "_endpoints_health_check")
 
                 finally:
                     # after the health check for that endpoint setting the timeouts back to their original values