File descriptors leak in undetected (UC) mode #3679

felipehertzer · 2025-04-18T03:38:46Z

felipehertzer
Apr 18, 2025

Hi,

I’m seeing a steady increase in open file descriptors whenever I activate UC mode (uc=True), leading eventually to:

filedescriptor out of range in select()

When UC mode is off the FD count remains flat.

When I use uc_open_with_reconnect and force a gc.collect() after teardown, the leak stops.

Below are three minimal repros—please let me know what else I can provide.

CDP

This script repeatedly opens a browser, calls activate_cdp_mode(...), then quits. You can see the number of FDs jump by ~8–10 each time, and connections in CLOSE_WAIT remain.

import os
import time

import psutil
from seleniumbase import SB


def _fd_exists(fd: int) -> bool:
    try:
        os.fstat(fd)
        return True
    except OSError:
        return False


def inspect_fds():
    p = psutil.Process()
    print(f"Count FD (psutil): {p.num_fds()}")
    connections = p.net_connections()
    if connections:
        print("Open connections:")
        for conn in connections:
            print(f"  {conn}")


if __name__ == "__main__":
    print("Start")
    inspect_fds()

    print('Browser Running')

    for i in range(1, 4):
        print(f"\nIteration {i}")
        with SB(uc=True) as browser:
            browser.activate_cdp_mode("https://example.com")
            browser.sleep(1)

        time.sleep(1)
        inspect_fds()

Output:

Start
Count FD (psutil): 3
Browser Running

Iteration 1
Count FD (psutil): 11
Open connections:
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51824), raddr=addr(ip='::1', port=51681), status='CLOSE_WAIT')
  pconn(fd=6, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51743), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=7, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51750), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')

Iteration 2
Count FD (psutil): 17
Open connections:
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51824), raddr=addr(ip='::1', port=51681), status='CLOSE_WAIT')
  pconn(fd=6, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51743), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=7, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51750), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=9, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51968), raddr=addr(ip='::1', port=51837), status='CLOSE_WAIT')
  pconn(fd=17, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51890), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=18, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51891), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')

Iteration 3
Count FD (psutil): 26
Open connections:
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51824), raddr=addr(ip='::1', port=51681), status='CLOSE_WAIT')
  pconn(fd=6, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51743), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=7, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51750), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=9, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51968), raddr=addr(ip='::1', port=51837), status='CLOSE_WAIT')
  pconn(fd=16, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=52119), raddr=addr(ip='::1', port=51976), status='CLOSE_WAIT')
  pconn(fd=17, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51890), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=18, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=51891), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=27, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=52043), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')
  pconn(fd=28, family=<AddressFamily.AF_INET: 2>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='127.0.0.1', port=52044), raddr=addr(ip='127.0.0.1', port=9222), status='CLOSE_WAIT')

UC without GC

Here we use uc_open_with_reconnect(...) instead of CDP. The leak still occurs, growing by ~1 FD per iteration.

import os
import time

import psutil
from seleniumbase import SB


def _fd_exists(fd: int) -> bool:
    try:
        os.fstat(fd)
        return True
    except OSError:
        return False


def inspect_fds():
    p = psutil.Process()
    print(f"Count FD (psutil): {p.num_fds()}")
    connections = p.net_connections()
    if connections:
        print("Open connections:")
        for conn in connections:
            print(f"  {conn}")


if __name__ == "__main__":
    print("Start")
    inspect_fds()

    print('Browser Running')

    for i in range(1, 10):
        print(f"\nIteration {i}")
        with SB(uc=True) as browser:
            browser.uc_open_with_reconnect("https://example.com", 4)

        time.sleep(1)
        inspect_fds()

Output:

Start
Count FD (psutil): 3
Browser Running

Iteration 1
Count FD (psutil): 9
Open connections:
  pconn(fd=3, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=50953), raddr=addr(ip='::1', port=50847), status='CLOSE_WAIT')

Iteration 2
Count FD (psutil): 10
Open connections:
  pconn(fd=3, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=50953), raddr=addr(ip='::1', port=50847), status='CLOSE_WAIT')
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51075), raddr=addr(ip='::1', port=50976), status='CLOSE_WAIT')

Iteration 3
Count FD (psutil): 11
Open connections:
  pconn(fd=3, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=50953), raddr=addr(ip='::1', port=50847), status='CLOSE_WAIT')
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51075), raddr=addr(ip='::1', port=50976), status='CLOSE_WAIT')
  pconn(fd=6, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=51192), raddr=addr(ip='::1', port=51090), status='CLOSE_WAIT')

UC using GC

Adding a manual garbage collection after each teardown appears to reclaim all sockets—FD count stays constant. But it only works for uc_open_with_reconnect.

import gc
import os
import time

import psutil
from seleniumbase import SB


def _fd_exists(fd: int) -> bool:
    try:
        os.fstat(fd)
        return True
    except OSError:
        return False


def inspect_fds():
    p = psutil.Process()
    print(f"Count FD (psutil): {p.num_fds()}")
    connections = p.net_connections()
    if connections:
        print("Open connections:")
        for conn in connections:
            print(f"  {conn}")


if __name__ == "__main__":
    print("Start")
    inspect_fds()

    print('Browser Running')

    for i in range(1, 4):
        print(f"\nIteration {i}")
        with SB(uc=True) as browser:
            browser.uc_open_with_reconnect("https://example.com", 4)

        time.sleep(1)
        gc.collect()
        inspect_fds()

Output:

Start
Count FD (psutil): 3
Browser Running

Iteration 1
Count FD (psutil): 6
Open connections:
  pconn(fd=3, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=49683), raddr=addr(ip='::1', port=49619), status='CLOSE_WAIT')

Iteration 2
Count FD (psutil): 6
Open connections:
  pconn(fd=5, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=49759), raddr=addr(ip='::1', port=49690), status='CLOSE_WAIT')

Iteration 3
Count FD (psutil): 6
Open connections:
  pconn(fd=3, family=<AddressFamily.AF_INET6: 30>, type=<SocketKind.SOCK_STREAM: 1>, laddr=addr(ip='::1', port=49996), raddr=addr(ip='::1', port=49776), status='CLOSE_WAIT')

Environment:
SeleniumBase version: e.g. 4.37.2
Python version: e.g. 3.13
OS: MacOS and Linux
Chrome: 135.0.7049.85
Chromedriver: matching Chrome

mdmintz · 2025-04-18T19:24:48Z

mdmintz
Apr 18, 2025
Maintainer

Do you know how to find specifically the code where the file descriptors leaks occur? The specific locations would be helpful. Much of UC Mode / CDP Mode is based on undetected-chromedriver / nodriver, which had several memory leaks: https://github.com/search?q=repo%3Aultrafunkamsterdam%2Fundetected-chromedriver+%22memory+leak%22&type=issues

0 replies

felipehertzer · 2025-04-19T01:40:54Z

felipehertzer
Apr 19, 2025
Author

Hello @mdmintz,

Thank you for your response. I spent several hours debugging and have identified a number of issues:

1. Session-based requests
I replaced all calls to requests.get with a with requests.Session() as session: context manager in both init.py and cdp.py.

Original

SeleniumBase/seleniumbase/undetected/__init__.py

Line 148 in b9a89f7

res = requests.get("http://127.0.0.1:9222", timeout=1)

SeleniumBase/seleniumbase/undetected/cdp.py

Line 104 in b9a89f7

resp = self._session.get(self.server_addr + uri)

Updated

https://github.com/felipehertzer/SeleniumBase/blob/f09035dcc138018d5cb208d6d65aba55929d64b7/seleniumbase/undetected/__init__.py#L139

https://github.com/felipehertzer/SeleniumBase/blob/f09035dcc138018d5cb208d6d65aba55929d64b7/seleniumbase/undetected/cdp.py#L132

You can review all of my changes here:
master...felipehertzer:SeleniumBase:fd-leak

https://github.com/felipehertzer/SeleniumBase/blob/f09035dcc138018d5cb208d6d65aba55929d64b7/seleniumbase/undetected/__init__.py#L570

With these adjustments, the leak in uc_open_with_reconnect no longer occurs (even without invoking the garbage collector), and the number of leaked file descriptors in activate_cdp_mode is greatly reduced.

2. Remaining leak in activate_cdp_mode
Despite the improvements above, two file descriptors (FD 6 and FD 7) still remain open when invoking activate_cdp_mode. Both originate from the start and get functions in seleniumbase/undetected/cdp_driver/browser.py:

SeleniumBase/seleniumbase/undetected/cdp_driver/browser.py

Line 303 in b9a89f7

async def start(self=None) -> Browser:

SeleniumBase/seleniumbase/undetected/cdp_driver/browser.py

Line 255 in b9a89f7

async def get(

It appears that each of these functions establishes its own socket connection, which is never closed. Moreover, I have confirmed that the corresponding stop method (at line 569) is never invoked, which I believe is the underlying cause of the remaining leak:

SeleniumBase/seleniumbase/undetected/cdp_driver/browser.py

Line 569 in b9a89f7

def stop(self):

Diagnostic output
Below is the output demonstrating the two unclosed sockets and their respective call stacks:

FD=6 and FD=7 are the two socks that do not close, and this is the trace of when each is called

[CONNECT] fd=6 → ('127.0.0.1', 9222)
    /test//.venv/lib/python3.13/site-packages/seleniumbase/fixtures/base_case.py:4881 in activate_cdp_mode
    /test//.venv/lib/python3.13/site-packages/seleniumbase/core/browser_launcher.py:5426 in <lambda>
    /test//.venv/lib/python3.13/site-packages/seleniumbase/core/browser_launcher.py:559 in uc_open_with_cdp_mode
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:712 in run_until_complete
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:683 in run_forever
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:2040 in _run_once
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/events.py:89 in _run
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/cdp_util.py:318 in start
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/browser.py:122 in create
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/browser.py:414 in start
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/connection.py:396 in send
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/connection.py:267 in aopen
    /test//.venv/lib/python3.13/site-packages/websockets/asyncio/client.py:541 in __await_impl__
    /test//.venv/lib/python3.13/site-packages/websockets/asyncio/client.py:467 in create_connection
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:1141 in create_connection
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:1044 in _connect_sock
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/selector_events.py:639 in sock_connect
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/selector_events.py:649 in _sock_connect
    /test//tests/test.py:181 in connect
    /test//tests/test.py:191 in _print_stack

[NEW SOCKET]  fd=7  family=<AddressFamily.AF_INET: 2>  type=<SocketKind.SOCK_STREAM: 1>
    /test//tests/test.py:205 in <module>
    /test//.venv/lib/python3.13/site-packages/seleniumbase/fixtures/base_case.py:4881 in activate_cdp_mode
    /test//.venv/lib/python3.13/site-packages/seleniumbase/core/browser_launcher.py:5426 in <lambda>
    /test//.venv/lib/python3.13/site-packages/seleniumbase/core/browser_launcher.py:607 in uc_open_with_cdp_mode
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:712 in run_until_complete
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:683 in run_forever
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:2040 in _run_once
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/events.py:89 in _run
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/browser.py:294 in get
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/connection.py:396 in send
    /test//.venv/lib/python3.13/site-packages/seleniumbase/undetected/cdp_driver/connection.py:267 in aopen
    /test//.venv/lib/python3.13/site-packages/websockets/asyncio/client.py:541 in __await_impl__
    /test//.venv/lib/python3.13/site-packages/websockets/asyncio/client.py:467 in create_connection
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:1141 in create_connection
    /opt/homebrew/Cellar/[email protected]/3.13.2/Frameworks/Python.framework/Versions/3.13/lib/python3.13/asyncio/base_events.py:1022 in _connect_sock
    /test//tests/test.py:175 in __init__
    /test//tests/test.py:191 in _print_stack

0 replies

mdmintz · 2025-04-19T15:58:16Z

mdmintz
Apr 19, 2025
Maintainer

4.37.3 - https://github.com/seleniumbase/SeleniumBase/releases/tag/v4.37.3
(At least UC Mode without CDP Mode no longer has memory leaks. Thank you!)

CDP Mode memory improvements will have to be figured out later.

0 replies

felipehertzer · 2025-04-20T08:24:31Z

felipehertzer
Apr 20, 2025
Author

Hey @mdmintz,

Thank you for the update.

After running some additional tests I’ve traced the file‑descriptor leak to the Connection class. Although an aclose() coroutine exists, it’s never invoked. The aexit() method would normally trigger aclose(), but it only executes when Connection is used as an async context manager—which isn’t happening in the current workflow.

To prevent the leak it will need to make sure every connection is explicitly closed, either by:

Refactoring usage to an async with Connection(...) as conn: pattern, or
Calling await self.aclose() at the end of send() (and any other method that opens a connection).

0 replies

mdmintz · 2025-04-20T15:27:21Z

mdmintz
Apr 20, 2025
Maintainer

Fixed in 4.37.4 - https://github.com/seleniumbase/SeleniumBase/releases/tag/v4.37.4

2 replies

mdmintz Apr 20, 2025
Maintainer

Thanks again for your help on this, @felipehertzer!

felipehertzer Apr 20, 2025
Author

That is great, it seems totally fixed, Thank you.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

File descriptors leak in undetected (UC) mode #3679

{{title}}

Replies: 5 comments 2 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

Select a reply

File descriptors leak in undetected (UC) mode #3679

felipehertzer Apr 18, 2025

CDP

UC without GC

UC using GC

Replies: 5 comments · 2 replies

mdmintz Apr 18, 2025 Maintainer

felipehertzer Apr 19, 2025 Author

mdmintz Apr 19, 2025 Maintainer

felipehertzer Apr 20, 2025 Author

mdmintz Apr 20, 2025 Maintainer

mdmintz Apr 20, 2025 Maintainer

felipehertzer Apr 20, 2025 Author

felipehertzer
Apr 18, 2025

Replies: 5 comments 2 replies

mdmintz
Apr 18, 2025
Maintainer

felipehertzer
Apr 19, 2025
Author

mdmintz
Apr 19, 2025
Maintainer

felipehertzer
Apr 20, 2025
Author

mdmintz
Apr 20, 2025
Maintainer

mdmintz Apr 20, 2025
Maintainer

felipehertzer Apr 20, 2025
Author