def setUp(self):
self.job = Job.create(FakeJobScheduler(),
self.launcher,
name='fake_job',
script_filename='fake_script',
stdout='fake_stdout',
stderr='fake_stderr',
sched_account='fake_account',
sched_partition='fake_partition',
sched_reservation='fake_reservation',
sched_nodelist="mynode",
sched_exclude_nodelist='fake_exclude_nodelist',
sched_exclusive_access='fake_exclude_access',
sched_options=['--fake'])
self.job.num_tasks = 4
self.job.num_tasks_per_node = 2
self.job.num_tasks_per_core = 1
self.job.num_tasks_per_socket = 1
self.job.num_cpus_per_task = 2
self.job.use_smt = True
self.job.time_limit = '10m'
self.job.options += ['--gres=gpu:4', '#DW jobdw anything']
self.job.launcher.options = ['--foo']
self.minimal_job = Job.create(FakeJobScheduler(),
self.launcher,
name='fake_job')
self.minimal_job.launcher.options = ['--foo']
def getallnodes(state='all', jobs_cli_options=None):
rt = runtime.runtime()
nodes = {}
for part in rt.system.partitions:
# This job will not be submitted, it's used only to filter
# the nodes based on the partition configuration
dummy_job = Job.create(part.scheduler,
part.launcher_type(),
name='placeholder-job',
sched_access=part.access,
sched_options=jobs_cli_options)
available_nodes = part.scheduler.allnodes()
available_nodes = part.scheduler.filternodes(dummy_job,
available_nodes)
getlogger().debug(
f'Total available nodes for {part.name}: {len(available_nodes)}'
)
if state.casefold() != 'all':
available_nodes = {n for n in available_nodes
if n.in_state(state)}
getlogger().debug(
f'[F] Selecting nodes in state {state!r}: '
f'available nodes now: {len(available_nodes)}'
)
nodes[part.fullname] = [n.name for n in available_nodes]
return nodes
def _remote_detect(part):
def _emit_script(job, env):
launcher_cmd = job.launcher.run_command(job)
commands = [
f'./bootstrap.sh',
f'{launcher_cmd} ./bin/reframe --detect-host-topology=topo.json'
]
job.prepare(commands, env, trap_errors=True)
getlogger().info(
f'Detecting topology of remote partition {part.fullname!r}: '
f'this may take some time...')
topo_info = {}
try:
prefix = runtime.runtime().get_option('general/0/remote_workdir')
with _copy_reframe(prefix) as dirname:
with osext.change_dir(dirname):
job = Job.create(part.scheduler,
part.launcher_type(),
name='rfm-detect-job',
sched_access=part.access)
_emit_script(job, [part.local_env])
getlogger().debug('submitting detection script')
_log_contents(job.script_filename)
job.submit()
job.wait()
getlogger().debug('job finished')
_log_contents(job.stdout)
_log_contents(job.stderr)
topo_info = json.loads(_contents('topo.json'))
except Exception as e:
getlogger().warning(f'failed to retrieve remote processor info: {e}')
return topo_info
def _setup_fake_check():
# A bit hacky, but we don't want to run a full test every time
test = _FakeCheck()
test._job = Job.create(
getscheduler('local')(),
getlauncher('local')(), 'fakejob')
test.job._completion_time = time.time()
return test
def _make_job(**jobargs):
return Job.create(scheduler(),
launcher(),
name='testjob',
workdir=tmp_path,
script_filename=str(tmp_path / 'job.sh'),
stdout=str(tmp_path / 'job.out'),
stderr=str(tmp_path / 'job.err'),
**jobargs)
def fake_check():
class _FakeCheck(rfm.RegressionTest):
pass
# A bit hacky, but we don't want to run a full test every time
test = _FakeCheck()
test._job = Job.create(
getscheduler('local')(),
getlauncher('local')(), 'fakejob')
test.job._completion_time = time.time()
return test
def _make_flexible_job(flex_type, **jobargs):
ret = Job.create(slurm_scheduler_patched,
getlauncher('local')(),
name='testjob',
workdir=tmp_path,
script_filename=str(tmp_path / 'job.sh'),
stdout=str(tmp_path / 'job.out'),
stderr=str(tmp_path / 'job.err'),
sched_flex_alloc_nodes=flex_type,
**jobargs)
ret.num_tasks = 0
ret.num_tasks_per_node = 4
return ret
def setUp(self):
self.workdir = tempfile.mkdtemp(dir='unittests')
self.testjob = Job.create(
self.scheduler,
self.launcher,
name='testjob',
workdir=self.workdir,
script_filename=os_ext.mkstemp_path(dir=self.workdir,
suffix='.sh'),
stdout=os_ext.mkstemp_path(dir=self.workdir, suffix='.out'),
stderr=os_ext.mkstemp_path(dir=self.workdir, suffix='.err'),
)
self.environs = [Environment(name='foo', modules=['testmod_foo'])]
self.pre_run = ['echo prerun']
self.post_run = ['echo postrun']
self.parallel_cmd = 'hostname'
def setUp(self):
# Monkey patch scheduler to simulate retrieval of nodes from Slurm
patched_sched = getscheduler('slurm')()
patched_sched.allnodes = self.create_dummy_nodes
patched_sched._get_default_partition = lambda: 'pdef'
self.workdir = tempfile.mkdtemp(dir='unittests')
self.testjob = Job.create(
patched_sched,
getlauncher('local')(),
name='testjob',
workdir=self.workdir,
script_filename=os.path.join(self.workdir, 'testjob.sh'),
stdout=os.path.join(self.workdir, 'testjob.out'),
stderr=os.path.join(self.workdir, 'testjob.err'))
self.testjob._sched_flex_alloc_nodes = 'all'
self.testjob.num_tasks_per_node = 4
self.testjob.num_tasks = 0
def fake_check():
class _FakeCheck(rfm.RegressionTest):
pass
@sn.deferrable
def error():
raise BaseException
# A bit hacky, but we don't want to run a full test every time
test = _FakeCheck()
test._job = Job.create(
getscheduler('local')(),
getlauncher('local')(), 'fakejob')
test.job._completion_time = time.time()
test.job._jobid = 12345
test.job._nodelist = ['localhost']
test.custom = 'hello extras'
test.custom_list = ['custom', 3.0, ['hello', 'world']]
test.custom_dict = {'a': 1, 'b': 2}
test.deferred = sn.defer('hello')
test.deferred_error = error()
return test
def fake_check():
class _FakeCheck(rfm.RegressionTest):
param = parameter(range(3), loggable=True, fmt=lambda x: 10 * x)
custom = variable(str, value='hello extras', loggable=True)
custom_list = variable(list,
value=['custom', 3.0, ['hello', 'world']],
loggable=True)
custom_dict = variable(dict, value={'a': 1, 'b': 2}, loggable=True)
# x is a variable that is loggable, but is left undefined. We want to
# make sure that logging does not crash and simply reports is as
# undefined
x = variable(str, loggable=True)
# A bit hacky, but we don't want to run a full test every time
test = _FakeCheck(variant_num=1)
test._job = Job.create(
getscheduler('local')(),
getlauncher('local')(), 'fakejob')
test.job._completion_time = time.time()
test.job._jobid = 12345
test.job._nodelist = ['localhost']
return test
def _make_job(launcher, *args, **kwargs):
return Job.create(FakeJobScheduler(), launcher, 'fake_job', *args,
**kwargs)
def make_job(self, *args, **kwargs):
return Job(*args, **kwargs)
