Uncaught exception - After upgrading python version

Hi,

I'm using Tornado 5.1.1
Recently I upgraded my python from 2.7.15 to 3.6.8.

Yesterday I got a strange exceptions for a few seconds (tons of that):

2020-04-04 21:04:39.153 ERROR [4592 140100505929536 tornado.general] Uncaught exception
Traceback (most recent call last):
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/http1connection.py", line 735, in _server_request_loop
    ret = yield conn.read_response(request_delegate)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/gen.py", line 1133, in run
    value = future.result()
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/gen.py", line 1147, in run
    yielded = self.gen.send(value)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/http1connection.py", line 245, in _read_message
    self.stream.set_close_callback(self._on_connection_close)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 606, in set_close_callback
    self._maybe_add_error_listener()
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 1128, in _maybe_add_error_listener
    self._add_io_state(ioloop.IOLoop.READ)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 1158, in _add_io_state
    self.fileno(), self._handle_events, self._state)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/platform/asyncio.py", line 76, in add_handler
    raise ValueError("fd %s added twice" % fd)
ValueError: fd 65 added twice
2020-04-04 21:04:39.417 ERROR [4118 140423800596288 tornado.general] Uncaught exception
Traceback (most recent call last):
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/http1connection.py", line 735, in _server_request_loop
    ret = yield conn.read_response(request_delegate)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/gen.py", line 1133, in run
    value = future.result()
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/gen.py", line 1147, in run
    yielded = self.gen.send(value)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/http1connection.py", line 245, in _read_message
    self.stream.set_close_callback(self._on_connection_close)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 606, in set_close_callback
    self._maybe_add_error_listener()
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 1128, in _maybe_add_error_listener
    self._add_io_state(ioloop.IOLoop.READ)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/iostream.py", line 1158, in _add_io_state
    self.fileno(), self._handle_events, self._state)
  File "/srv/dy/rcomserver/venv/lib64/python3.6/site-packages/tornado/platform/asyncio.py", line 76, in add_handler
    raise ValueError("fd %s added twice" % fd)
ValueError: fd 63 added twice

Any idea why it happened? or how to debug/reproduce? not sure if related to my upgrade or not but I never saw that before..

Start with the traceback locations in tornado/http1connection.py, tornado/iostream.py, and tornado/platform/asyncio.py, then inspect how file descriptors are registered during the server request loop. Try to reproduce the duplicate fd registration across the reported Python versions. Done means identifying the cause and capturing the behavior in a regression test or clearly documenting why it cannot be reproduced.