PYTHON-1838 Discard ServerSessions involved in network errors

Author: ShaneHarvey

pymongo/client_session.py

@@ -786,6 +786,15 @@ def __init__(self):
         self.session_id = {'id': Binary(uuid.uuid4().bytes, 4)}
         self.last_use = monotonic.time()
         self._transaction_id = 0
+        self.dirty = False
+
+    def mark_dirty(self):
+        """Mark this session as dirty.
+
+        A server session is marked dirty when a command fails with a network
+        error. Dirty sessions are later discarded from the server session pool.
+        """
+        self.dirty = True
 
     def timed_out(self, session_timeout_minutes):
         idle_seconds = monotonic.time() - self.last_use
@@ -832,10 +841,11 @@ def get_server_session(self, session_timeout_minutes):
     def return_server_session(self, server_session, session_timeout_minutes):
         self._clear_stale(session_timeout_minutes)
         if not server_session.timed_out(session_timeout_minutes):
-            self.appendleft(server_session)
+            self.return_server_session_no_lock(server_session)
 
     def return_server_session_no_lock(self, server_session):
-        self.appendleft(server_session)
+        if not server_session.dirty:
+            self.appendleft(server_session)
 
     def _clear_stale(self, session_timeout_minutes):
         # Clear stale sessions. The least recently used are on the right.

pymongo/mongo_client.py

@@ -1312,6 +1312,8 @@ def _reset_on_error(self, server_address, session):
             # Server Discovery And Monitoring Spec: "When an application
             # operation fails because of any network error besides a socket
             # timeout...."
+            if session:
+                session._server_session.mark_dirty()
             raise
         except NotMasterError:
             # "When the client sees a "not master" error it MUST replace the
@@ -1323,6 +1325,8 @@ def _reset_on_error(self, server_address, session):
             # "Client MUST replace the server's description with type Unknown
             # ... MUST NOT request an immediate check of the server."
             self.__reset_server(server_address)
+            if session:
+                session._server_session.mark_dirty()
             raise
         except OperationFailure as exc:
             if exc.code in helpers._RETRYABLE_ERROR_CODES: