fix launch port already in use

python/paddle/distributed/fleet/launch_utils.py

@@ -12,9 +12,7 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-import functools
 import logging
-import socket
 import time
 import os
 import signal
@@ -27,6 +25,7 @@
 import socket
 import warnings
 import six
+import struct
 
 import paddle
 import paddle.fluid as fluid
@@ -362,6 +361,10 @@ def add_arguments(argname, type, default, help, argparser, **kwargs):
 def find_free_ports(num):
     def __free_port():
         with closing(socket.socket(socket.AF_INET, socket.SOCK_STREAM)) as s:
+            # Note(wangxi): Close the connection with a TCP RST instead
+            # of a TCP FIN, to avoid time_wait state.
+            s.setsockopt(socket.SOL_SOCKET, socket.SO_LINGER,
+                         struct.pack('ii', 1, 0))
             s.bind(('', 0))
             return s.getsockname()[1]
 
@@ -376,7 +379,7 @@ def __free_port():
             return port_set
 
         step += 1
-        if step > 100:
+        if step > 400:
             print(
                 "can't find avilable port and use the specified static port now!"
             )

python/paddle/fluid/tests/unittests/test_launch_coverage.py

@@ -24,6 +24,7 @@
 
 from argparse import ArgumentParser, REMAINDER
 from paddle.distributed.utils import _print_arguments, get_gpus, get_cluster_from_args
+from paddle.distributed.fleet.launch_utils import find_free_ports
 
 
 def _parse_args():
@@ -115,6 +116,9 @@ def test_gpus(self):
         args.use_paddlecloud = True
         cluster, pod = get_cluster_from_args(args, "0")
 
+    def test_find_free_ports(self):
+        find_free_ports(2)
+
 
 if __name__ == '__main__':
     unittest.main()