There appear to be 1 leaked semaphore objects to clean up at shutdown

[RabiyaF]

• CodeCarbon version: 2.8.0
• Python version: 3.10.15
• Operating System: Mac OS (Sequoia)

Description

We were using codecarbon 2.6.0 for fitbenchmarking and this was working fine.

We then updated to 2.7.1 and were seeing the"EmissionsTracker" object has no attribute '_scheduler' error.

Today, I updated the codecarbon dependency to test the latest version 2.8.0. Everything seems to work okay. But if I do a KeyboardInterrupt, I see the following warning:

UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '

If I downgrade to codecarbon 2.6.0 and do a KeyboardInterrupt I do not see this warning. I see the normal stack trace of the KeyboardInterrupt.

Question: Has something been updated recently that might be causing this warning ?

[benoit-cty]

Hello,
Thanks for reporting this and using CodeCarbon.

It may have been introduced by https://github.com/mlco2/codecarbon/pull/701/files
Maybe it is because the scheduler is not stopped properly in this case ? @inimaz do you have an idea ?

But I don't see the message you mention when I do a Ctrl+C, can you tell me more on how you get it ?

[inimaz]

Hi!
Thanks for using Codecarbon!
Indeed as @benoit-cty suggested it could be this feature. Could you provide a snippet of how you are using codecarbon + some more logs? Otherwise I think it is hard to tell.
Thanks!

[RabiyaF]

We are using codecarbon v2.6.0 like this

tracker = EmissionsTracker(measure_power_secs=1)
tracker.start_task()
"""
our computations go here
"""
energy = tracker.stop_task().energy_consumed

With v2.6.0 I see this in the logs when I do a KeyboardInterrupt. I think this is the stack trace that I would expect with a KeyboardInterrupt.

Running NIST average_difficulty problem set with scipy minimizers 

Running the benchmarking on the NIST_average_difficulty problem set
Password:
Parsing problems
Running problems
                                                                                                                                                                                                        
#################################
 Running data from: ENSO.dat 1/9                                                                                                                                                                        
#################################                                                                                                                                                                       
    Starting value: 1/2                                                                                                                                                                                 
        Cost Function: weighted_nlls                                                                                                                                                                    
            Software: SCIPY                                                                                                                                                                             
                Minimizer: Nelder-Mead                                                                                                                                                                  
Captured output:                                                                                                                                                                                        
ref: /Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/data/private_infra/global_energy_mix.json                                                                   
ref: /Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/data/private_infra/global_energy_mix.json

                Minimizer: Powell                                                                                                                                                                       
Benchmark problems:   0%|                                                                                                                                          | 0/9 [00:06<?, ?Benchmark problem/s^Captured output:                                                                                                                                                                                        
ref: /Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/data/private_infra/global_energy_mix.json                                                                   

Benchmark problems:   0%|                                                                                                                                          | 0/9 [00:08<?, ?Benchmark problem/s]
Traceback (most recent call last):                                                                                                                                                                      
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/bin/fitbenchmarking", line 8, in <module>
    sys.exit(main())
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/cli/main.py", line 562, in main
    run(
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/cli/exception_handler.py", line 29, in wrapped
    return f(*args, **kwargs)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/cli/main.py", line 410, in run
    results, failed_problems, unselected_minimizers = fit.benchmark()
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 144, in benchmark
    results = self.__loop_over_starting_values(problem)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 203, in __loop_over_starting_values
    result = self.__loop_over_cost_function(problem)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 243, in __loop_over_cost_function
    result = self.__loop_over_fitting_software(cost_func)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 297, in __loop_over_fitting_software
    result, minimizer_failed = self.__loop_over_minimizers(
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 364, in __loop_over_minimizers
    result = self.__loop_over_jacobians(controller)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 416, in __loop_over_jacobians
    result = self.__loop_over_hessians(controller)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 480, in __loop_over_hessians
    accuracy, runtimes, energy = self.__perform_fit(controller)
  File "/Users/<usrpath>/Desktop/Github/Work/fitbenchmarking/fitbenchmarking/core/fitting_benchmarking.py", line 533, in __perform_fit
    energy = tracker.stop_task().energy_consumed / num_runs
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/emissions_tracker.py", line 503, in stop_task
    self._measure_power_and_energy()
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/emissions_tracker.py", line 758, in _measure_power_and_energy
    self._do_measurements()
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/emissions_tracker.py", line 735, in _do_measurements
    f"{hardware.__class__.__name__} : {hardware.total_power().W:,.2f} "
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/external/hardware.py", line 455, in total_power
    return self._get_power()
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/external/hardware.py", line 427, in _get_power
    all_details: Dict = self._interface.get_details()
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/core/powermetrics.py", line 143, in get_details
    self._log_values()
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/site-packages/codecarbon/core/powermetrics.py", line 126, in _log_values
    returncode = subprocess.call(cmd, universal_newlines=True)
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/subprocess.py", line 347, in call
    return p.wait(timeout=timeout)
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/subprocess.py", line 1209, in wait
    return self._wait(timeout=timeout)
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/subprocess.py", line 1959, in _wait
    (pid, sts) = self._try_wait(0)
  File "/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/subprocess.py", line 1917, in _try_wait
    (pid, sts) = os.waitpid(self.pid, wait_flags)
KeyboardInterrupt

With v2.8.0 I see this in the logs when I do a KeyboardInterrupt

Running NIST average_difficulty problem set with scipy minimizers 

Running the benchmarking on the NIST_average_difficulty problem set
Password:
Parsing problems
Running problems
                                                                                                                                                                                                        
#################################
 Running data from: ENSO.dat 1/9                                                                                                                                                                        
#################################                                                                                                                                                                       
    Starting value: 1/2                                                                                                                                                                                 
        Cost Function: weighted_nlls                                                                                                                                                                    
            Software: SCIPY                                                                                                                                                                             
                Minimizer: Nelder-Mead                                                                                                                                                                  
                Minimizer: Powell                                                                                                                                                                       
^C%
/Users/<usrpath>/mambaforge/envs/mantid_env/lib/python3.10/multiprocessing/resource_tracker.py:224: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '

Let me know if any other information is required. I will try to see if it is something from our codebase that is causing this issue.