diff --git a/sql/python/sqlflow_submitter/db.py b/sql/python/sqlflow_submitter/db.py
index 511138015f..756587f16a 100644
--- a/sql/python/sqlflow_submitter/db.py
+++ b/sql/python/sqlflow_submitter/db.py
@@ -17,7 +17,7 @@
 import tensorflow as tf
 import sqlflow_submitter.db_writer as db_writer
 
-def connect(driver, database, user, password, host, port, auth=""):
+def connect(driver, database, user, password, host, port, session_cfg={}, auth=""):
     if driver == "mysql":
         # NOTE: use MySQLdb to avoid bugs like infinite reading:
         # https://bugs.mysql.com/bug.php?id=91971
@@ -32,19 +32,21 @@ def connect(driver, database, user, password, host, port, auth=""):
         return connect(database)
     elif driver == "hive":
         from impala.dbapi import connect
-        return connect(user=user,
+        conn = connect(user=user,
                        password=password,
                        database=database,
                        host=host,
                        port=int(port),
                        auth_mechanism=auth)
+        conn.session_cfg = session_cfg
+        return conn
     elif driver == "maxcompute":
         from sqlflow_submitter.maxcompute import MaxCompute
         return MaxCompute.connect(database, user, password, host)
 
     raise ValueError("unrecognized database driver: %s" % driver)
 
-def db_generator(driver, conn, session_cfg, statement,
+def db_generator(driver, conn, statement,
                  feature_column_names, label_column_name,
                  feature_specs, fetch_size=128):
     def read_feature(raw_val, feature_spec, feature_name):
@@ -69,7 +71,7 @@ def read_feature(raw_val, feature_spec, feature_name):
 
     def reader():
         if driver == "hive":
-            cursor = conn.cursor(configuration=session_cfg)
+            cursor = conn.cursor(configuration=conn.session_cfg)
         else:
             cursor = conn.cursor()
         cursor.execute(statement)
diff --git a/sql/python/sqlflow_submitter/db_test.py b/sql/python/sqlflow_submitter/db_test.py
index f3b9861cd4..0f142320c0 100644
--- a/sql/python/sqlflow_submitter/db_test.py
+++ b/sql/python/sqlflow_submitter/db_test.py
@@ -132,7 +132,7 @@ def test_generator(self):
                 "is_sparse": False,
                 "shape": []
             }}
-            gen = db_generator(driver, conn, {}, "SELECT * FROM test_table_float_fea",
+            gen = db_generator(driver, conn, "SELECT * FROM test_table_float_fea",
                                ["features"], "label", column_name_to_type)
             idx = 0
             for d in gen():
@@ -159,6 +159,6 @@ def test_generate_fetch_size(self):
                 }}
             
 
-            gen = db_generator(driver, conn, {}, 'SELECT * FROM iris.train limit 10',
+            gen = db_generator(driver, conn, 'SELECT * FROM iris.train limit 10',
                                 ["sepal_length"], "class", column_name_to_type, fetch_size=4)
             self.assertEqual(len([g for g in gen()]), 10)
diff --git a/sql/template_analyze.go b/sql/template_analyze.go
index 752fb84599..29746c13d0 100644
--- a/sql/template_analyze.go
+++ b/sql/template_analyze.go
@@ -51,10 +51,10 @@ session_cfg = {}
 session_cfg["{{$k}}"] = "{{$v}}"
 {{end}}
 
-conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}")
+conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}",session_cfg=session_cfg)
 
 def analyzer_dataset():
-    stream = db_generator(driver, conn, session_cfg, """{{.AnalyzeDatasetSQL}}""", feature_names, label_name, feature_metas)
+    stream = db_generator(driver, conn, """{{.AnalyzeDatasetSQL}}""", feature_names, label_name, feature_metas)
     xs = pd.DataFrame(columns=feature_names)
     ys = pd.DataFrame(columns=[label_name])
     i = 0
diff --git a/sql/template_tf.go b/sql/template_tf.go
index 990acdfa0e..a7b36118ab 100644
--- a/sql/template_tf.go
+++ b/sql/template_tf.go
@@ -42,7 +42,12 @@ database=""
 database="{{.Database}}"
 {{end}}
 
-conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}")
+session_cfg = {}
+{{ range $k, $v := .Session }}
+session_cfg["{{$k}}"] = "{{$v}}"
+{{end}}
+
+conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}",session_cfg=session_cfg)
 
 feature_column_names = [{{range .X}}
 "{{.FeatureName}}",
@@ -70,11 +75,6 @@ feature_metas["{{$value.FeatureName}}"] = {
 }
 {{end}}
 
-session_cfg = {}
-{{ range $k, $v := .Session }}
-session_cfg["{{$k}}"] = "{{$v}}"
-{{end}}
-
 def get_dtype(type_str):
     if type_str == "float32":
         return tf.float32
@@ -104,7 +104,7 @@ def input_fn(datasetStr):
         else:
             feature_types.append(get_dtype(feature_metas[name]["dtype"]))
 
-    gen = db_generator(driver, conn, session_cfg, datasetStr, feature_column_names, "{{.Y.FeatureName}}", feature_metas)
+    gen = db_generator(driver, conn, datasetStr, feature_column_names, "{{.Y.FeatureName}}", feature_metas)
     dataset = tf.data.Dataset.from_generator(gen, (tuple(feature_types), tf.{{.Y.Dtype}}))
     ds_mapper = functools.partial(_parse_sparse_feature, feature_metas=feature_metas)
     return dataset.map(ds_mapper)
@@ -169,7 +169,12 @@ database="{{.Database}}"
 database=""
 {{end}}
 
-conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}")
+session_cfg = {}
+{{ range $k, $v := .Session }}
+session_cfg["{{$k}}"] = "{{$v}}"
+{{end}}
+
+conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}",session_cfg=session_cfg)
 
 feature_column_names = [{{range .X}}
 "{{.FeatureName}}",
@@ -197,11 +202,6 @@ feature_metas["{{$value.FeatureName}}"] = {
 }
 {{end}}
 
-session_cfg = {}
-{{ range $k, $v := .Session }}
-session_cfg["{{$k}}"] = "{{$v}}"
-{{end}}
-
 def get_dtype(type_str):
     if type_str == "float32":
         return tf.float32
@@ -232,7 +232,7 @@ def eval_input_fn(batch_size):
         else:
             feature_types.append(get_dtype(feature_metas[name]["dtype"]))
 
-    gen = db_generator(driver, conn, session_cfg, """{{.PredictionDatasetSQL}}""",
+    gen = db_generator(driver, conn, """{{.PredictionDatasetSQL}}""",
         feature_column_names, "{{.Y.FeatureName}}", feature_metas)
     dataset = tf.data.Dataset.from_generator(gen, (tuple(feature_types), tf.{{.Y.Dtype}}))
     ds_mapper = functools.partial(_parse_sparse_feature, feature_metas=feature_metas)
@@ -322,7 +322,7 @@ class FastPredict:
 
 column_names = feature_column_names[:]
 column_names.append("{{.Y.FeatureName}}")
-pred_gen = db_generator(driver, conn, session_cfg, """{{.PredictionDatasetSQL}}""", feature_column_names, "{{.Y.FeatureName}}", feature_metas)()
+pred_gen = db_generator(driver, conn, """{{.PredictionDatasetSQL}}""", feature_column_names, "{{.Y.FeatureName}}", feature_metas)()
 fast_predictor = FastPredict(classifier, fast_input_fn)
 
 with buffered_db_writer(driver, conn, "{{.TableName}}", column_names, 100) as w:
diff --git a/sql/template_xgboost.go b/sql/template_xgboost.go
index 8ad022ada2..dc4ceee25a 100644
--- a/sql/template_xgboost.go
+++ b/sql/template_xgboost.go
@@ -58,10 +58,10 @@ feature_specs["{{$value.FeatureName}}"] = {
 }
 {{end}}
 
-conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}")
+conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}",session_cfg=session_cfg)
 
 def xgb_dataset(fn, dataset_sql):
-    gen = db_generator(driver, conn, session_cfg, dataset_sql, feature_column_names, "{{.Y.FeatureName}}", feature_specs)
+    gen = db_generator(driver, conn, dataset_sql, feature_column_names, "{{.Y.FeatureName}}", feature_specs)
     with open(fn, 'w') as f:
         for item in gen():
             features, label = item
@@ -117,10 +117,10 @@ feature_specs["{{$value.FeatureName}}"] = {
 }
 {{end}}
 
-conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}")
+conn = connect(driver, database, user="{{.User}}", password="{{.Password}}", host="{{.Host}}", port={{.Port}}, auth="{{.Auth}}",session_cfg=session_cfg)
 
 def xgb_dataset(fn, dataset_sql):
-    gen = db_generator(driver, conn, session_cfg, dataset_sql, feature_column_names, "", feature_specs)
+    gen = db_generator(driver, conn, dataset_sql, feature_column_names, "", feature_specs)
     with open(fn, 'w') as f:
         for item in gen():
             features, label = item