PYTHON-525 Don't reset whole Cluster on network error.

ajdavis · ajdavis · commit fd3154d8a1ff · 2014-09-09T13:49:57.000-04:00
According to the spec, clear the pool for the server we were disconnected from
and mark the server Unknown, but don't update other servers. This is a change
from PyMongo 2's MongoReplicaSetClient, which reset all servers and their pools
when disconnected from the primary specifically.

The spec's justification is, "since application sockets are used frequently, a
network error likely means the server has just become unavailable, so an
immediate refresh is likely to get a network error, too. The server will not
remain Unknown forever. It will be refreshed by the next periodic check or, if
an application operation needs the server sooner than that, then a re-check
will be triggered by the server selection algorithm."

If the primary has gone down and we are in the midst of an election, marking
all servers Unknown and forcing a re-scan will simply detect that all servers
are currently secondaries, which the client knows anyway.
diff --git a/pymongo/mongo_client.py b/pymongo/mongo_client.py
@@ -810,22 +810,17 @@ def _send_message_with_response(
             exhaust)
 
     def _reset_on_error(self, server, fn, *args, **kwargs):
-        """Execute an operation. Reset the pool on network error.
+        """Execute an operation. Reset the server on network error.
 
         Returns fn()'s return value on success. On error, clears the server's
-        pool and marks the server Unknown or, if the server is the primary,
-        resets all pools and servers.
+        pool and marks the server Unknown.
 
         Re-raises any exception thrown by fn().
         """
         try:
             return fn(*args, **kwargs)
         except ConnectionFailure:
-            if server.description.is_writable:
-                self.disconnect()
-            else:
-                self._cluster.reset_server(server.address)
-
+            self._cluster.reset_server(server.description.address)
             raise
 
     def start_request(self):
diff --git a/test/test_client.py b/test/test_client.py
@@ -39,6 +39,7 @@
                             OperationFailure,
                             CursorNotFound)
 from pymongo.server_selectors import writable_server_selector
+from pymongo.server_type import SERVER_TYPE
 from test import (client_context,
                   client_knobs,
                   connection_string,
@@ -1109,6 +1110,48 @@ def test_reconnect(self):
         self.assertEqual('a', c.host)
         self.assertEqual(1, c.port)
 
+    def test_network_error_on_operation(self):
+        # Verify only the disconnected server is reset by a network failure.
+
+        # Disable background refresh.
+        with client_knobs(heartbeat_frequency=999999):
+            c = MockClient(
+                standalones=[],
+                members=['a:1', 'b:2'],
+                mongoses=[],
+                host='a:1',
+                replicaSet='rs',
+                connect=False)
+
+            # Set host-specific information so we can test whether it is reset.
+            c.set_wire_version_range('a:1', 0, 1)
+            c.set_wire_version_range('b:2', 0, 2)
+
+            connected(c)
+            wait_until(lambda: len(c.nodes) == 2, 'connect')
+
+            sd = c._get_cluster().get_server_by_address(('a', 1)).description
+            self.assertEqual(SERVER_TYPE.RSPrimary, sd.server_type)
+            self.assertEqual(0, sd.min_wire_version)
+            self.assertEqual(1, sd.max_wire_version)
+
+            c.kill_host('a:1')
+
+            # MongoClient is disconnected from the primary.
+            self.assertRaises(AutoReconnect, c.db.collection.find_one)
+
+            # The primary's description is reset.
+            sd_a = c._get_cluster().get_server_by_address(('a', 1)).description
+            self.assertEqual(SERVER_TYPE.Unknown, sd_a.server_type)
+            self.assertEqual(0, sd_a.min_wire_version)
+            self.assertEqual(0, sd_a.max_wire_version)
+
+            # ...but not the secondary's.
+            sd_b = c._get_cluster().get_server_by_address(('b', 2)).description
+            self.assertEqual(SERVER_TYPE.RSSecondary, sd_b.server_type)
+            self.assertEqual(0, sd_b.min_wire_version)
+            self.assertEqual(2, sd_b.max_wire_version)
+
 
 if __name__ == "__main__":
     unittest.main()
