diff --git a/python/pipeline/channel.py b/python/pipeline/channel.py
index 8a03c30b9075aa56f88d8f876932cf6a68835c62..b04714a5efece5d2acff3dc9e041147cfeb910ba 100644
--- a/python/pipeline/channel.py
+++ b/python/pipeline/channel.py
@@ -330,8 +330,8 @@ class ProcessChannel(object):
 
     def front(self, op_name=None, timeout=None):
         _LOGGER.debug(
-            self._log("{} try to get data[?]; timeout={}".format(op_name,
-                                                                 timeout)))
+            self._log("{} try to get data[?]; timeout(s)={}".format(op_name,
+                                                                    timeout)))
         endtime = None
         if timeout is not None:
             if timeout <= 0:
@@ -603,8 +603,8 @@ class ThreadChannel(Queue.Queue):
 
     def front(self, op_name=None, timeout=None):
         _LOGGER.debug(
-            self._log("{} try to get data[?]; timeout={}".format(op_name,
-                                                                 timeout)))
+            self._log("{} try to get data[?]; timeout(s)={}".format(op_name,
+                                                                    timeout)))
         endtime = None
         if timeout is not None:
             if timeout <= 0:
diff --git a/python/pipeline/operator.py b/python/pipeline/operator.py
index eea153d4beb019956e429af61af6e2e13476e0e0..2728c6e7382e9d8d0364c29ec4fe85adb5054a70 100644
--- a/python/pipeline/operator.py
+++ b/python/pipeline/operator.py
@@ -87,8 +87,9 @@ class Op(object):
                          .format(self.name, self._batch_size))
             self._batch_size = 1
         if self._auto_batching_timeout != None:
-            _LOGGER.warn("Op({}) reset auto_batching_timeout=1 (original: {})"
-                         .format(self.name, self._auto_batching_timeout))
+            _LOGGER.warn(
+                "Op({}) reset auto_batching_timeout=None (original: {})"
+                .format(self.name, self._auto_batching_timeout))
             self._auto_batching_timeout = None
 
     def use_profiler(self, use_profile):
@@ -235,6 +236,9 @@ class Op(object):
                 target=self._run,
                 args=(concurrency_idx, self._get_input_channel(),
                       self._get_output_channels(), client_type, True))
+            # When a process exits, it attempts to terminate
+            # all of its daemonic child processes.
+            t.daemon = True
             t.start()
             threads.append(t)
         return threads
@@ -310,7 +314,7 @@ class Op(object):
                             _LOGGER.error(error_info)
                         else:
                             _LOGGER.warn(
-                                log_func("timeout, retry({}/{})"
+                                log_func("PaddleService timeout, retry({}/{})"
                                          .format(i + 1, self._retry)))
                     except Exception as e:
                         ecode = ChannelDataEcode.UNKNOW.value
@@ -399,7 +403,7 @@ class Op(object):
         while True:
             batch = []
             _LOGGER.debug(
-                log_func("Auto-batching expect size: {}; timeout: {}".format(
+                log_func("Auto-batching expect size: {}; timeout(s): {}".format(
                     batch_size, timeout)))
             while len(batch) == 0:
                 endtime = None