Orca: add shutdown API to tf2 pyspark estimator. (#5499)

* feat: add shutdown API to tf2 pyspark estimator.

* fix: replace raise with log4Error.

* refactor: refactor the stop log server function.

* fix: fix code style issue.

python/orca/src/bigdl/orca/learn/log_monitor.py

@@ -164,3 +164,30 @@ def _print_logs():
     logger_thread.daemon = True
     logger_thread.start()
     return logger_thread
+
+
+def stop_log_server(thread, ip, port):
+    if thread.is_alive():
+        import inspect
+        import ctypes
+        import zmq
+
+        def _async_raise(tid, exctype):
+            tid = ctypes.c_long(tid)
+            if not inspect.isclass(exctype):
+                exctype = type(exctype)
+            res = ctypes.pythonapi.PyThreadState_SetAsyncExc(tid, ctypes.py_object(exctype))
+            if res == 0:
+                invalidInputError(False, "invalid thread id")
+            elif res != 1:
+                ctypes.pythonapi.PyThreadState_SetAsyncExc(tid, None)
+                invalidOperationError(False, "PyThreadState_SetAsyncExc failed")
+
+        def stop_thread(thread):
+            _async_raise(thread.ident, SystemExit)
+
+        context = zmq.Context()
+        socket = context.socket(zmq.REQ)
+        socket.connect("tcp://{}:{}".format(ip, port))
+        socket.send_string("shutdown log server")
+        stop_thread(thread)

python/orca/src/bigdl/orca/learn/tf2/pyspark_estimator.py

@@ -38,7 +38,7 @@
 from bigdl.orca.learn.utils import maybe_dataframe_to_xshards, dataframe_to_xshards, \
     convert_predict_xshards_to_dataframe, make_data_creator, load_model, \
     save_model, process_xshards_of_pandas_dataframe
-from bigdl.orca.learn.log_monitor import start_log_server
+from bigdl.orca.learn.log_monitor import start_log_server, stop_log_server
 from bigdl.orca.data.shard import SparkXShards
 from bigdl.orca import OrcaContext
 from bigdl.dllib.utils.log4Error import invalidInputError
@@ -94,7 +94,7 @@ def __init__(self,
         is_local = sc.master.startswith("local")
         self.need_to_log_to_driver = (not is_local) and log_to_driver
         if self.need_to_log_to_driver:
-            start_log_server(self.ip, self.port)
+            self.log_server_thread = start_log_server(self.ip, self.port)
 
     def _get_cluster_info(self, sc):
         cluster_info = self.workerRDD.barrier().mapPartitions(find_ip_and_free_port).collect()
@@ -528,3 +528,9 @@ def get_model(self):
     @property
     def _model_saved_path(self):
         return os.path.join(self.model_dir, "{}_model.h5".format(self.application_id))
+
+    def shutdown(self):
+        """
+        Shutdown estimator and release resources.
+        """
+        stop_log_server(self.log_server_thread, self.ip, self.port)