def test_slurmenv_api_cancel_nl():
if not in_slurm_allocation() or get_num_slurm_nodes() < 2:
pytest.skip('test not run in slurm allocation or allocation is smaller than 2 nodes')
resources, allocation = get_slurm_resources_binded()
set_pythonpath_to_qcg_module()
tmpdir = str(tempfile.mkdtemp(dir=SHARED_PATH))
print(f'tmpdir: {tmpdir}')
try:
m = LocalManager(['--log', 'debug', '--wd', tmpdir, '--report-format', 'json'], {'wdir': str(tmpdir)})
iters=10
ids = m.submit(Jobs().
add(exec='/bin/sleep', args=['5s'], iteration=iters, stdout='sleep.out.${it}',
stderr='sleep.err.${it}', numCores=1)
)
jid = ids[0]
assert len(m.list()) == 1
list_jid = list(m.list().keys())[0]
assert list_jid == jid
# wait for job to start executing
sleep(2)
m.cancel([jid])
m.wait4(m.list())
jinfos = m.info_parsed(ids, withChilds=True)
assert all((len(jinfos) == 1, jid in jinfos, jinfos[jid].status == 'CANCELED'))
# the canceled iterations are included in 'failed' entry in job statistics
# the cancel status is presented in 'childs/state' entry
assert all((jinfos[jid].iterations, jinfos[jid].iterations.get('start', -1) == 0,
jinfos[jid].iterations.get('stop', 0) == iters, jinfos[jid].iterations.get('total', 0) == iters,
jinfos[jid].iterations.get('finished', 0) == iters, jinfos[jid].iterations.get('failed', -1) == iters))
assert len(jinfos[jid].childs) == iters
for iteration in range(iters):
job_it = jinfos[jid].childs[iteration]
assert all((job_it.iteration == iteration, job_it.name == '{}:{}'.format(jid, iteration),
job_it.status == 'CANCELED')), str(job_it)
m.remove(jid)
finally:
m.finish()
m.cleanup()
rmtree(tmpdir)
def test_local_manager_submit_simple(tmpdir):
cores = 4
# switch on debugging (by default in api.log file)
m = LocalManager(['--wd', str(tmpdir), '--nodes',
str(cores)], {'wdir': str(tmpdir)})
try:
res = m.resources()
assert all(
('total_nodes' in res, 'total_cores'
in res, res['total_nodes'] == 1, res['total_cores'] == cores))
ids = m.submit(Jobs().add(name='host',
exec='/bin/hostname',
args=['--fqdn'],
stdout='host.stdout').add(
name='date',
exec='/bin/date',
stdout='date.stdout',
numCores={'exact': 2}))
assert len(m.list()) == 2
m.wait4(ids)
jinfos = m.info(ids)
assert all(
('jobs' in jinfos, len(jinfos['jobs'].keys()) == 2, 'host'
in jinfos['jobs'], 'date' in jinfos['jobs'],
jinfos['jobs']['host'].get('data', {}).get('status',
'') == 'SUCCEED',
jinfos['jobs']['date'].get('data', {}).get('status',
'') == 'SUCCEED'))
aux_dir = find_single_aux_dir(str(tmpdir))
assert all(
(exists(tmpdir.join('.qcgpjm-client', 'api.log')),
exists(join(aux_dir,
'service.log')), exists(tmpdir.join('host.stdout')),
exists(tmpdir.join('date.stdout'))))
finally:
m.finish()
# m.stopManager()
m.cleanup()
def test_resume_simple(tmpdir):
try:
ncores = 4
m = LocalManager(['--log', 'debug', '--wd', tmpdir, '--report-format', 'json', '--nodes', str(ncores)],
{'wdir': str(tmpdir)})
its = 10
job_req = {
'name': 'sleep',
'execution': {
'exec': '/bin/sleep',
'args': [ '4s' ],
'stdout': 'out',
},
'iteration': { 'stop': its },
'resources': { 'numCores': { 'exact': 1 } }
}
jobs = Jobs().add_std(job_req)
job_ids = m.submit(jobs)
# because job iterations executes in order, after finish of 4th iteration, the three previous should also finish
m.wait4('sleep:3')
jinfos = m.info_parsed(job_ids, withChilds=True)
assert jinfos
jinfo = jinfos['sleep']
assert all((jinfo.iterations, jinfo.iterations.get('start', -1) == 0,
jinfo.iterations.get('stop', 0) == its, jinfo.iterations.get('total', 0) == its,
jinfo.iterations.get('finished', 0) == ncores, jinfo.iterations.get('failed', -1) == 0)), str(jinfo)
assert len(jinfo.childs) == its
for iteration in range(its):
job_it = jinfo.childs[iteration]
exp_status = ['SUCCEED']
if iteration > 3:
exp_status = ['EXECUTING', 'SCHEDULED', 'QUEUED']
assert all((job_it.iteration == iteration,
job_it.name == '{}:{}'.format('sleep', iteration),
job_it.status in exp_status)),\
f"{job_it.iteration} != {iteration}, {job_it.name} != {'{}:{}'.format('sleep', iteration)}, {job_it.status} != {exp_status}"
# kill process
m.kill_manager_process()
m.cleanup()
ncores = 4
m = LocalManager(['--log', 'debug', '--wd', tmpdir, '--report-format', 'json', '--nodes', str(ncores),
'--resume', tmpdir],
{'wdir': str(tmpdir)})
m.wait4all()
jinfos = m.info_parsed(job_ids, withChilds=True)
assert jinfos
jinfo = jinfos['sleep']
assert all((jinfo.iterations, jinfo.iterations.get('start', -1) == 0,
jinfo.iterations.get('stop', 0) == its, jinfo.iterations.get('total', 0) == its,
jinfo.iterations.get('finished', 0) == its, jinfo.iterations.get('failed', -1) == 0)), str(jinfo)
assert len(jinfo.childs) == its
for iteration in range(its):
job_it = jinfo.childs[iteration]
assert all((job_it.iteration == iteration,
job_it.name == '{}:{}'.format('sleep', iteration),
job_it.status == 'SUCCEED')), \
f"{job_it.iteration} != {iteration}, {job_it.name} != {'{}:{}'.format('sleep', iteration)}, {job_it.status} != SUCCEED"
finally:
if m:
m.finish()
m.cleanup()
cmd = '%s run_mscale.py --submodel %s --data_dir=%s --instance_index %d --coupling_type %s --num_instances %d --weather_coupling %s' % (
PYTHON_CMD, SUBMODEL, DATA_DIR, INSTANCE_INDEX, COUPLING_TYPE,
NUM_INSTANCES, WEATHER_COUPLING)
print("\tAdd job with cmd = %s" % (cmd))
TaskID = 'TaskID%d_%s' % (INSTANCE_INDEX + 1, SUBMODEL)
stderr = 'log_task/%s_${jname}__${uniq}.stderr' % (TaskID)
stdout = 'log_task/%s_${jname}__${uniq}.stdout' % (TaskID)
jobs.add(name=TaskID,
exec='bash',
args=['-c', cmd],
stdout=stdout,
stderr=stderr,
numCores={'exact': INSTANCE_CORES},
model='default')
INSTANCE_INDEX = INSTANCE_INDEX + 1
ids = m.submit(jobs)
# wait until submited jobs finish
m.wait4(ids)
# get detailed information about submited and finished jobs
print("jobs details:\n%s\n" % str(m.info(ids)))
m.finish()
m.kill_manager_process()
m.cleanup()