`pycriu`: `intermittent issue` when using `use_binary`

[herheliuk]

When using pycriu with use_binary after a reinstall, I'm getting an error on the first run (only):

pycriu.criu.CRIUExceptionExternal: DUMP failed: Unexpected response type 0: Error(0): Unknown

It is noteworthy that there are no logs in the dump folder.

Steps to reproduce using Ubuntu

Prepare

mkdir ./pycriu-use-binary-bug/ && cd ./pycriu-use-binary-bug/
sudo apt update && sudo apt install -y python3-venv

Reinstall

deactivate
sudo rm -rf * .*
git clone https://github.com/checkpoint-restore/criu ./criu/
sudo bash ./criu/contrib/dependencies/apt-packages.sh
sudo make -C ./criu/ install
python3 -m venv ./env/
source ./env/bin/activate
pip install ./criu/lib/

Adding scripts

test.py

#!/usr/bin/env python3

dumps_dir = './criu_dumps/'

from pycriu import criu as pycriu_criu
from shutil import which as shutil_which
from os import (
    open as os_open,
    close as os_close,
    O_DIRECTORY as os_O_DIRECTORY,
    makedirs as os_makedirs
)
from contextlib import contextmanager

@contextmanager
def open_directory_fd(dir_path: str):
    fd = os_open(dir_path, os_O_DIRECTORY)
    try:
        yield fd
    finally:
        os_close(fd)

criu_binary = shutil_which('criu')

print(f'{dumps_dir=}')
print(f'{criu_binary=}\n')

criu = pycriu_criu()
criu.use_binary(criu_binary)

pid = int(input('Please enter pid to dump: '))

criu.opts.pid = pid
criu.opts.shell_job = True

os_makedirs(dumps_dir, exist_ok=True)

with open_directory_fd(dumps_dir) as fd:
    criu.opts.images_dir_fd = fd
    criu.dump()

print(f'\nSuccessfully dumped {pid}.')

process_to_dump.py

#!/usr/bin/env python3

from os import getpid as os_getpid
from time import sleep as time_sleep

print(f'pid {os_getpid()}\n')

i = 0
    
while i < 100:
    i += 1
    print(i, end=' ', flush=True)
    time_sleep(1)

Run

Terminal 2

source ./env/bin/activate
sudo -E $(which python) test.py

Terminal 3

tmux new-session 'python3 process_to_dump.py'

Notes

If not seeing an error, try again following the steps from Reinstall,
Usually shows up at least every second run, I suggest trying up to 3 times.

Expected behaviour

No errors in terminal 1
& killed process in terminal 2

Same behaviour is expected on recurring runs (scripts can run multiple times).

[herheliuk]

It would've been nice if pycriu was giving sane errors instead of 'code 0' every time 😅

But work in progress I guess, maybe even I would find how to fix that ... who knows.

[herheliuk]

Potential reason

Encountered when dumping for the 'first time' using CLI

Warn  (criu/kerndat.c:1250): Stale /run/criu.kdat file

[herheliuk]

I am surprised the tests are okay with it, there's probably no tests for this
or the issue is actually shows up starting from a second install...

[github-actions]

A friendly reminder that this issue had no activity for 30 days.