BOP solver unresponsive, disrespects time limit

[snedeljkovic]

What version of OR-tools and what language are you using?
Version: 7.5.7466
Language: Python

Which solver are you using (e.g. CP-SAT, Routing Solver, GLOP, BOP, Gurobi)
BOP

What operating system (Linux, Windows, ...) and version?
Ubuntu 18.04.3 LTS (Google Colab)

What did you do?
Steps to reproduce the behavior:
I set the time limit to 5 min.

What did you expect to see
I expected the solver to terminate after ~5 min, or respond to a cell interrupt command.

What did you see instead?
The solver doesn't terminate after the time limit, and is unresponsive to interrupt. I had to manually restart the kernel.

Anything else we should know about your project / environment
I ran it on Google Colab. Sometimes ~2x smaller instances exhibit the same behavior although I cannot reproduce it consistently. I attached the model that fails every time.
model_lp.zip

[Mizux]

AFAIK, since the solver is in C++ and we only swig wrap it in python, signal are not forwarded to the underlying libraries solver.

For timeout not respected need to find some information internally....

[snedeljkovic]

Is this a confirmed bug (on your side), and if so will it be added to a milestone? Can I somehow further help you?

[snedeljkovic]

@Mizux @lperron I built the master branch from source and the model still fails.

[jundl77]

Is there any update on this? I have also noticed that ortools is not responding to signals, which is a problem especially if it is the SIGTERM signal

[lperron]

Use the SAT solver. It answers to SIGTERM and honor the time limit :-) Laurent Perron | Operations Research | lperron@google.com | (33) 1 42 68 53 00 Le mar. 15 déc. 2020 à 12:22, Julian Brendl <notifications@github.com> a écrit :

…

Is there any update on this? I have also noticed that ortools is not responding to signals, which is a problem especially if it is the SIGTERM signal — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1886 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACUPL3PRQRKG3FCNELH7EJ3SU5BI3ANCNFSM4KXET7LA> .

[jundl77]

Thanks for your quick reply, but I found that not to be true. I have created a little example below that demonstrates this.

As soon as the code enters the "start solving" step, i.e. the search for solutions begins, the CP-SAT solver also no longer replies to SIGTERM. You can see this is if you run this in your terminal and press Ctrl+C once that "start solving" line has been printed.

Here is the code that reproduces this, it's just a very simple solver that solves the subset-sum problem:

from ortools.sat.python import cp_model
import random


class VarArraySolutionPrinter(cp_model.CpSolverSolutionCallback):
    """Print intermediate solutions."""

    def __init__(self, variables):
        cp_model.CpSolverSolutionCallback.__init__(self)
        self.__variables = variables
        self.__solution_count = 0

    def on_solution_callback(self):
        self.__solution_count += 1
        for v in self.__variables:
            print('%i' % (self.Value(v)), end=' ')
        print()

    def solution_count(self):
        return self.__solution_count


def SearchForAllSolutionsSampleSat(values, total_sum):
    """Showcases calling the solver to search for all solutions."""
    # Creates the model.
    model = cp_model.CpModel()

    num_values = len(values)
    subsets = [model.NewIntVar(0, num_values, f'{i}') for i in range(num_values)]
    subset_sum = model.NewIntVar(0, num_values, 'sum')

    # add constraints on each variable to ensure it is only used at most once
    [model.Add(subset <= 1) for subset in subsets]
    model.Add(subset_sum == cp_model.LinearExpr.Sum(subsets))
    model.Add(total_sum == cp_model.LinearExpr.ScalProd(subsets, values))

    # create a solver and solve.
    solver = cp_model.CpSolver()
    solution_printer = VarArraySolutionPrinter(subsets)
    print('start solving')
    status = solver.SearchForAllSolutions(model, solution_printer)

    print('Status = %s' % solver.StatusName(status))
    print('Number of solutions found: %i' % solution_printer.solution_count())

print('generate data')
s = [random.randint(1, 100) for _ in range(2000000)]

print('enter solver')
SearchForAllSolutionsSampleSat(s, 452343)