use AI_NUMERICHOST | AI_NUMERICSERV to skip getaddrinfo thread in asyncio

[graingert]

BPO	46824
Nosy	@asvetlov, @bdarnell, @1st1, @graingert
PRs	gh-90980: skip getaddrinfo thread if host is already resolved, using socket.AI_NUMERIC... #31497

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = None
closed_at = None
created_at = <Date 2022-02-22.10:15:28.937>
labels = ['type-feature', 'expert-asyncio']
title = 'use AI_NUMERICHOST | AI_NUMERICSERV to skip getaddrinfo thread in asyncio'
updated_at = <Date 2022-03-18.16:45:34.609>
user = 'https://github.com/graingert'

bugs.python.org fields:

activity = <Date 2022-03-18.16:45:34.609>
actor = 'Ben.Darnell'
assignee = 'none'
closed = False
closed_date = None
closer = None
components = ['asyncio']
creation = <Date 2022-02-22.10:15:28.937>
creator = 'graingert'
dependencies = []
files = []
hgrepos = []
issue_num = 46824
keywords = ['patch']
message_count = 5.0
messages = ['413699', '413701', '413702', '415510', '415511']
nosy_count = 4.0
nosy_names = ['asvetlov', 'Ben.Darnell', 'yselivanov', 'graingert']
pr_nums = ['31497']
priority = 'normal'
resolution = None
stage = 'patch review'
status = 'open'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue46824'
versions = []

[graingert]

now that the getaddrinfo lock has been removed on all platforms the numeric only host resolve in asyncio could be moved back into BaseEventLoop.getaddrinfo

[asvetlov]

Could you provide more context for the proposed change?

[graingert]

hello, it's actually a bit of a round about context, but it was brought up on a tornado issue where I was attempting to port the asyncio optimization to tornado: tornadoweb/tornado#3113 (comment)

I think it would be better to use this AI_NUMERICHOST | AI_NUMERICSERV optimization from trio everywhere instead

[bdarnell]

To summarize the justification, this patch does two things: it moves an optimization from create_connection to getaddrinfo, which makes it apply to more callers (including Tornado), and it makes the code simpler and less redundant (net reduction of 47 non-test lines in the patch).

As far as we can tell, the reason it wasn't done this way in the first place is that at the time getaddrinfo held a global lock on some platforms, but this is no longer true. If there's still some locking in or around getaddrinfo on some platforms (or some libc implementations), this patch would be a bad idea. Is there a good way to test for that? I suppose we could set up a deliberately-slow DNS server and try to call getaddrinfo with AI_NUMERICHOST while another thread is blocked talking to that server, but that seems like a lot of test infrastructure to build out.

[bdarnell]

On MacOS in 2015, getaddrinfo was found to be much slower than inet_pton. Unless that's changed, this patch would be a performance regression on that platform. Data and benchmark script in https://groups.google.com/g/python-tulip/c/-SFI8kkQEj4/m/m1-oCMSABgAJ

[graingert]

running the benchmark script https://gist.github.com/graingert/5097c6be997ab2a201b48b858524e163 on my Linux machine I get

before:

/home/graingert/projects/cpython-main/../demo/demo.py:28: DeprecationWarning: There is no current event loop
  loop = get_event_loop()
      host       port     family       type      proto    secs
   1.2.3.4          1          2          1          6     9.4
   1.2.3.4          1          0          1          6     9.7
   1.2.3.4          1          0          2         17    10.9
   1.2.3.4          1          0          1          0    14.5
   1.2.3.4          1          0          2          0    15.9
   1.2.3.4          1          0          0          0    15.8
       ::3          1         10          1          6    13.0
       ::3          1          0          1          6    13.3
Traceback (most recent call last):
  File "/home/graingert/projects/cpython-main/../demo/demo.py", line 58, in <module>
    lookup(*info)
    ^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython-main/../demo/demo.py", line 32, in lookup
    return loop.run_until_complete(loop.getaddrinfo(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython-main/Lib/asyncio/base_events.py", line 650, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython-main/Lib/asyncio/base_events.py", line 864, in getaddrinfo
    return await self.run_in_executor(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython-main/Lib/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython-main/Lib/socket.py", line 961, in getaddrinfo
    for res in _socket.getaddrinfo(host, port, family, type, proto, flags):
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
socket.gaierror: [Errno -2] Name or service not known

after:

/home/graingert/projects/cpython/../demo/demo.py:28: DeprecationWarning: There is no current event loop
  loop = get_event_loop()
      host       port     family       type      proto    secs
   1.2.3.4          1          2          1          6     2.4
   1.2.3.4          1          0          1          6     2.4
   1.2.3.4          1          0          2         17     2.4
   1.2.3.4          1          0          1          0     2.4
   1.2.3.4          1          0          2          0     2.3
   1.2.3.4          1          0          0          0     2.7
       ::3          1         10          1          6     2.3
       ::3          1          0          1          6     2.3
Traceback (most recent call last):
  File "/home/graingert/projects/cpython/../demo/demo.py", line 58, in <module>
    lookup(*info)
    ^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython/../demo/demo.py", line 32, in lookup
    return loop.run_until_complete(loop.getaddrinfo(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython/Lib/asyncio/base_events.py", line 591, in run_until_complete
    return future.result()
           ^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython/Lib/asyncio/base_events.py", line 814, in getaddrinfo
    return await self.run_in_executor(
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython/Lib/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/graingert/projects/cpython/Lib/socket.py", line 962, in getaddrinfo
    for res in _socket.getaddrinfo(host, port, family, type, proto, flags):
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
socket.gaierror: [Errno -2] Name or service not known

[graingert]

on macos:

before https://github.com/graingert/cpython/runs/6528334512?check_suite_focus=true#step:7:1

Run ./python.exe demo.py
/Users/runner/work/cpython/cpython/demo.py:28: DeprecationWarning: There is no current event loop
  loop = get_event_loop()
      host       port     family       type      proto    secs
   1.2.3.4          1          2          1          6    54.6
   1.2.3.4          1          0          1          6    55.0
   1.2.3.4          1          0          2         17    52.8
   1.2.3.4          1          0          1          0    52.0
   1.2.3.4          1          0          2          0    51.2
   1.2.3.4          1          0          0          0    54.7
       ::3          1         30          1          6    52.6
       ::3          1          0          1          6    52.6
   ::3%lo0          1         30          1          6    55.5

after https://github.com/graingert/cpython/runs/6528182360?check_suite_focus=true#step:7:11

Run ./python.exe demo.py
/Users/runner/work/cpython/cpython/demo.py:28: DeprecationWarning: There is no current event loop
  loop = get_event_loop()
      host       port     family       type      proto    secs
   1.2.3.4          1          2          1          6    21.3
   1.2.3.4          1          0          1          6    21.1
   1.2.3.4          1          0          2         17    21.1
   1.2.3.4          1          0          1          0    21.1
   1.2.3.4          1          0          2          0    21.0
   1.2.3.4          1          0          0          0    22.1
       ::3          1         30          1          6    20.9
       ::3          1          0          1          6    20.9
   ::3%lo0          1         30          1          6    23.8

[graingert]

https://bugzilla.mozilla.org/show_bug.cgi?id=404399#c16 and https://bugzilla.mozilla.org/show_bug.cgi?id=344809#c21 seem relevant

[graingert]

benchmarking _ipaddr_info graingert@dee834c seems to show that it's about the same speed as getaddrinfo with AI_NUMERICHOST:

https://github.com/graingert/cpython/runs/6528895181?check_suite_focus=true

Run ./python.exe demo.py
/Users/runner/work/cpython/cpython/demo.py:28: DeprecationWarning: There is no current event loop
  loop = get_event_loop()
      host       port     family       type      proto    secs
   1.2.3.4          1          2          1          6    19.2
   1.2.3.4          1          0          1          6    19.1
   1.2.3.4          1          0          2         17    19.2
   1.2.3.4          1          0          1          0    19.0
   1.2.3.4          1          0          2          0    18.9
   1.2.3.4          1          0          0          0    18.1
       ::3          1         30          1          6    19.0
       ::3          1          0          1          6    20.2
   ::3%lo0          1         30          1          6    19.1

@bdarnell