fix: Fix pickling lock errors in McpSessionManager

When we added the session_lock_map, it resulted in pickling errors during Agent Engine deployment. To fix, we implemented custom getstate and setstate methods to exclude the lock map lock and session lock map from pickling. Closes https://github.com/google/adk-python/issues/4486.

Co-authored-by: Kathy Wu <wukathy@google.com>
PiperOrigin-RevId: 871056554

tests/unittests/tools/mcp_tool/test_mcp_session_manager.py

@@ -607,6 +607,39 @@ class TestMCPSessionManager:
     exit_stack2.aclose.assert_not_called()
     assert len(manager._sessions) == 0
 
+  @pytest.mark.asyncio
+  async def test_pickle_mcp_session_manager(self):
+    """Verify that MCPSessionManager can be pickled and unpickled."""
+    import pickle
+
+    manager = MCPSessionManager(self.mock_stdio_connection_params)
+
+    # Access the lock to ensure it's initialized
+    lock = manager._session_lock
+    assert isinstance(lock, asyncio.Lock)
+
+    # Add a mock session to verify it's cleared on pickling
+    manager._sessions["test"] = (Mock(), Mock(), asyncio.get_running_loop())
+
+    # Pickle and unpickle
+    pickled = pickle.dumps(manager)
+    unpickled = pickle.loads(pickled)
+
+    # Verify basics are restored
+    assert unpickled._connection_params == manager._connection_params
+
+    # Verify transient/unpicklable members are re-initialized or cleared
+    assert unpickled._sessions == {}
+    assert unpickled._session_lock_map == {}
+    assert isinstance(unpickled._lock_map_lock, type(manager._lock_map_lock))
+    assert unpickled._lock_map_lock is not manager._lock_map_lock
+    assert unpickled._errlog == sys.stderr
+
+    # Verify we can still get a lock in the new instance
+    new_lock = unpickled._session_lock
+    assert isinstance(new_lock, asyncio.Lock)
+    assert new_lock is not lock
+
 
 @pytest.mark.asyncio
 async def test_retry_on_errors_decorator():