Beispiel #1
0
    def test_job_with_error_is_handled_correctly(self):
        # Given.
        problem = EllipticalDrop(self.sim_dir, self.output_dir)
        problem.cases[0].base_command += ' --xxx'
        s = self._make_scheduler()
        t = TaskRunner(tasks=[SolveProblem(problem=problem)], scheduler=s)

        # When.
        try:
            t.run(wait=1)
        except RuntimeError:
            pass

        # Then.

        # Ensure that the directories are copied over when they have errors.
        sim1 = os.path.join(self.root, self.sim_dir, 'elliptical_drop',
                            'no_update_h')
        self.assertTrue(os.path.exists(sim1))
        sim2 = os.path.join(self.root, self.sim_dir, 'elliptical_drop',
                            'update_h')
        self.assertTrue(os.path.exists(sim2))

        # Ensure that all the correct but already scheduled jobs are completed.
        task_status = t.task_status
        status_values = list(task_status.values())
        self.assertEqual(status_values.count('error'), 1)
        self.assertEqual(status_values.count('done'), 1)
        self.assertEqual(status_values.count('not started'), 1)
        for t, s in task_status.items():
            if s == 'done':
                self.assertTrue(t.complete())
            if s == 'error':
                self.assertRaises(RuntimeError, t.complete)
Beispiel #2
0
    def test_tasks_with_dependencies(self):
        # Given
        s = self._make_scheduler()
        cmd = 'python -c "import time; print(time.time())"'
        ct1_dir = os.path.join(self.sim_dir, '1')
        ct2_dir = os.path.join(self.sim_dir, '2')
        ct3_dir = os.path.join(self.sim_dir, '3')
        ct1 = CommandTask(cmd, output_dir=ct1_dir)
        ct2 = CommandTask(cmd, output_dir=ct2_dir, depends=[ct1])
        ct3 = CommandTask(cmd, output_dir=ct3_dir, depends=[ct1, ct2])

        # When
        t = TaskRunner(tasks=[ct1, ct2, ct3], scheduler=s)

        # Then
        self.assertEqual(len(t.todo), 3)

        # When
        t.run(wait=0.1)

        wait_until(lambda: not ct3.complete())

        # Then.
        # Ensure that the tasks are run in the right order.
        ct1_t, ct2_t, ct3_t = [
            self._get_time(x) for x in (ct1_dir, ct2_dir, ct3_dir)
        ]
        self.assertTrue(ct2_t > ct1_t)
        self.assertTrue(ct3_t > ct2_t)
Beispiel #3
0
    def test_task_runner_checks_for_error_in_running_tasks(self, m_t_cores):
        # Given
        s = self._make_scheduler()
        cmd = 'python -c "import sys, time; time.sleep(0.1); sys.exit(1)"'
        ct1_dir = os.path.join(self.sim_dir, '1')
        ct2_dir = os.path.join(self.sim_dir, '2')
        ct3_dir = os.path.join(self.sim_dir, '3')
        job_info = dict(n_core=2, n_thread=2)
        ct1 = CommandTask(cmd, output_dir=ct1_dir, job_info=job_info)
        ct2 = CommandTask(cmd, output_dir=ct2_dir, job_info=job_info)
        ct3 = CommandTask(cmd, output_dir=ct3_dir, job_info=job_info)

        # When
        t = TaskRunner(tasks=[ct1, ct2, ct3], scheduler=s)

        # Then
        self.assertEqual(len(t.todo), 3)

        # When
        n_errors = t.run(wait=0.1)

        # Then
        # In this case, two tasks should have run and one should not have run
        # as the other two had errors.
        self.assertEqual(n_errors, 2)
        self.assertEqual(len(t.todo), 1)
        self.assertTrue(os.path.exists(ct3_dir))
        self.assertTrue(os.path.exists(ct2_dir))
        self.assertFalse(os.path.exists(ct1_dir))
Beispiel #4
0
    def test_task_runner_waits_for_tasks_in_the_end(self, m_t_cores):
        # Given
        s = self._make_scheduler()
        cmd = 'python -c "import sys, time; time.sleep(0.1); sys.exit(1)"'
        ct1_dir = os.path.join(self.sim_dir, '1')
        ct2_dir = os.path.join(self.sim_dir, '2')
        ct3_dir = os.path.join(self.sim_dir, '3')
        ct1 = CommandTask(cmd, output_dir=ct1_dir)
        ct2 = CommandTask(cmd, output_dir=ct2_dir)
        ct3 = CommandTask(cmd, output_dir=ct3_dir)

        # When
        t = TaskRunner(tasks=[ct1, ct2, ct3], scheduler=s)
        n_errors = t.run(wait=0.1)

        # Then
        # All the tasks may have been run but those that ran will fail.
        self.assertEqual(n_errors + len(t.todo), 3)
        self.assertTrue(n_errors > 0)
Beispiel #5
0
    def test_nothing_is_run_when_output_exists(self):
        # Given.
        s = self._make_scheduler()
        output = os.path.join(self.output_dir, 'elliptical_drop')
        os.makedirs(output)

        # When
        problem = EllipticalDrop(self.sim_dir, self.output_dir)
        t = TaskRunner(tasks=[SolveProblem(problem=problem)], scheduler=s)

        # Then.
        self.assertEqual(len(t.todo), 0)
Beispiel #6
0
    def test_task_runner_does_not_add_repeated_tasks(self):
        # Given
        s = self._make_scheduler()
        cmd = 'python -c "print(1)"'
        ct1 = CommandTask(cmd, output_dir=self.sim_dir)
        ct2 = CommandTask(cmd, output_dir=self.sim_dir)

        # When
        t = TaskRunner(tasks=[ct1, ct2, ct1], scheduler=s)

        # Then
        self.assertEqual(len(t.todo), 1)
Beispiel #7
0
    def test_simulation_with_dependencies(self):
        # Given
        class A(Problem):
            def setup(self):
                cmd = 'python -c "import time; print(time.time())"'
                s1 = Simulation(self.input_path('1'), cmd)
                s2 = Simulation(self.input_path('2'), cmd, depends=[s1])
                s3 = Simulation(self.input_path('3'), cmd, depends=[s1, s2])
                self.cases = [s1, s2, s3]

            def run(self):
                self.make_output_dir()

        s = self._make_scheduler()

        # When
        problem = A(self.sim_dir, self.output_dir)
        task = SolveProblem(problem)
        t = TaskRunner(tasks=[task], scheduler=s)

        # Then
        self.assertEqual(len(t.todo), 4)
        # Basically only one instance of CommandTask should be created.
        names = [x.__class__.__name__ for x in t.todo]
        self.assertEqual(names.count('CommandTask'), 3)
        self.assertEqual(names.count('SolveProblem'), 1)

        # When
        t.run(wait=0.1)
        wait_until(lambda: not task.complete())

        # Then
        ct1_t, ct2_t, ct3_t = [
            self._get_time(problem.input_path(x)) for x in ('1', '2', '3')
        ]
        self.assertTrue(ct2_t > ct1_t)
        self.assertTrue(ct3_t > ct2_t)
Beispiel #8
0
    def test_automation(self):
        # Given.
        problem = EllipticalDrop(self.sim_dir, self.output_dir)
        s = self._make_scheduler()
        t = TaskRunner(tasks=[SolveProblem(problem=problem)], scheduler=s)

        # When.
        t.run(wait=1)

        # Then.
        sim1 = os.path.join(self.root, self.sim_dir, 'elliptical_drop',
                            'no_update_h')
        self.assertTrue(os.path.exists(sim1))
        sim2 = os.path.join(self.root, self.sim_dir, 'elliptical_drop',
                            'update_h')
        self.assertTrue(os.path.exists(sim2))

        results = os.path.join(self.root, self.output_dir, 'elliptical_drop',
                               'result.txt')
        self.assertTrue(os.path.exists(results))
        data = open(results).read()
        self.assertTrue('no_update_h' in data)
        self.assertTrue('update_h' in data)

        # When.
        problem = EllipticalDrop(self.sim_dir, self.output_dir)
        t = TaskRunner(tasks=[SolveProblem(problem=problem)], scheduler=s)

        # Then.
        self.assertEqual(len(t.todo), 0)

        # When
        problem.clean()

        # Then.
        out1 = os.path.join(self.root, self.output_dir, 'elliptical_drop',
                            'update_h')
        out2 = os.path.join(self.root, self.output_dir, 'elliptical_drop',
                            'no_update_h')
        self.assertFalse(os.path.exists(out1))
        self.assertFalse(os.path.exists(out2))
        self.assertTrue(os.path.exists(sim1))
        self.assertTrue(os.path.exists(sim2))
Beispiel #9
0
    def test_task_runner_doesnt_block_on_problem_with_error(self, m_t_cores):
        # Given
        class A(Problem):
            def get_requires(self):
                cmd = ('python -c "import sys, time; time.sleep(0.1); '
                       'sys.exit(1)"')
                ct = CommandTask(cmd, output_dir=self.input_path())
                return [('task1', ct)]

            def run(self):
                self.make_output_dir()

        s = self._make_scheduler()

        # When
        task = RunAll(simulation_dir=self.sim_dir,
                      output_dir=self.output_dir,
                      problem_classes=[A])
        t = TaskRunner(tasks=[task], scheduler=s)
        n_error = t.run(wait=0.1)

        # Then
        self.assertTrue(n_error > 0)

        # For the case where there is a match expression
        # When
        problem = A(self.sim_dir, self.output_dir)
        problem.clean()

        task = RunAll(simulation_dir=self.sim_dir,
                      output_dir=self.output_dir,
                      problem_classes=[A],
                      match='*task1')
        t = TaskRunner(tasks=[task], scheduler=s)
        n_error = t.run(wait=0.1)

        # Then
        self.assertTrue(n_error > 0)
Beispiel #10
0
    def test_problem_depending_on_other_problems(self):
        # Given
        class A(Problem):
            def get_requires(self):
                cmd = 'python -c "print(1)"'
                # Can return tasks ...
                ct = CommandTask(cmd, output_dir=self.sim_dir)
                return [('task1', ct)]

            def run(self):
                self.make_output_dir()

        class B(Problem):
            def get_requires(self):
                # or return Problem instances ...
                return [('a', A(self.sim_dir, self.out_dir))]

            def run(self):
                self.make_output_dir()

        class C(Problem):
            def get_requires(self):
                # ... or Problem subclasses
                return [('a', A), ('b', B)]

            def run(self):
                self.make_output_dir()

        s = self._make_scheduler()

        # When
        task = RunAll(simulation_dir=self.sim_dir,
                      output_dir=self.output_dir,
                      problem_classes=[A, B, C])
        t = TaskRunner(tasks=[task], scheduler=s)

        # Then
        self.assertEqual(len(t.todo), 5)
        # Basically only one instance of CommandTask should be created.
        names = [x.__class__.__name__ for x in t.todo]
        problems = [x.problem for x in t.todo if isinstance(x, SolveProblem)]
        self.assertEqual(names.count('RunAll'), 1)
        self.assertEqual(names.count('CommandTask'), 1)
        self.assertEqual(names.count('SolveProblem'), 3)
        self.assertEqual(len(problems), 3)
        self.assertEqual(sorted(x.__class__.__name__ for x in problems),
                         ['A', 'B', 'C'])

        # When
        t.run(wait=0.1)

        # Then.
        self.assertEqual(t.todo, [])

        # The output dirs should exist now.
        for name in ('A', 'B', 'C'):
            self.assertTrue(os.path.exists(os.path.join(self.output_dir,
                                                        name)))

        # When
        # We set force to True.
        task = RunAll(simulation_dir=self.sim_dir,
                      output_dir=self.output_dir,
                      problem_classes=[A, B, C],
                      force=True)

        # Then
        # Make sure that all the output directories are deleted as this will
        for name in ('A', 'B', 'C'):
            self.assertFalse(
                os.path.exists(os.path.join(self.output_dir, name)))