class BaseExecutor(threading.Thread, PluginFactory):
def __init__(self, **kwargs):
super(BaseExecutor, self).__init__()
self.setName("BaseExecutor")
self.queue = None
self.payload = None
self.args = None
for key in kwargs:
setattr(self, key, kwargs[key])
self.__stop = threading.Event()
self.__event_ranges = []
self.__is_set_payload = False
self.__is_retrieve_payload = False
self.communication_manager = None
self.proc = None
def get_pid(self):
return self.proc.pid if self.proc else None
def __del__(self):
self.stop()
if self.communication_manager:
self.communication_manager.stop()
def is_payload_started(self):
return False
def start(self):
super(BaseExecutor, self).start()
self.communication_manager = CommunicationManager()
self.communication_manager.start()
def stop(self):
if not self.is_stop():
self.__stop.set()
def is_stop(self):
return self.__stop.is_set()
def stop_communicator(self):
logger.info("Stopping communication manager")
if self.communication_manager:
while self.communication_manager.is_alive():
if not self.communication_manager.is_stop():
self.communication_manager.stop()
logger.info("Communication manager stopped")
def set_payload(self, payload):
self.payload = payload
self.__is_set_payload = True
job = self.get_job()
if job and job.workdir:
os.chdir(job.workdir)
def is_set_payload(self):
return self.__is_set_payload
def set_retrieve_payload(self):
self.__is_retrieve_payload = True
def is_retrieve_payload(self):
return self.__is_retrieve_payload
def retrieve_payload(self):
logger.info("Retrieving payload: %s" % self.args)
jobs = self.communication_manager.get_jobs(njobs=1, args=self.args)
logger.info("Received jobs: %s" % jobs)
if jobs:
job = create_job(jobs[0], queue=self.queue)
# get the payload command from the user specific code
pilot_user = os.environ.get('PILOT_USER', 'atlas').lower()
user = __import__('pilot.user.%s.common' % pilot_user, globals(),
locals(), [pilot_user], 0) # Python 2/3
cmd = user.get_payload_command(job)
logger.info("payload execution command: %s" % cmd)
payload = {'executable': cmd, 'workdir': job.workdir, 'job': job}
logger.info("Retrieved payload: %s" % payload)
return payload
return None
def get_payload(self):
if self.__is_set_payload:
return self.payload
def get_job(self):
return self.payload['job'] if self.payload and 'job' in list(
self.payload.keys()) else None # Python 2/3
def get_event_ranges(self, num_event_ranges=1, queue_factor=2):
if config.Payload.executor_type.lower() == 'raythena':
old_queue_factor = queue_factor
queue_factor = 1
logger.info("raythena - Changing queue_factor from %s to %s" %
(old_queue_factor, queue_factor))
logger.info(
"Getting event ranges: (num_ranges: %s) (queue_factor: %s)" %
(num_event_ranges, queue_factor))
if len(self.__event_ranges) < num_event_ranges:
ret = self.communication_manager.get_event_ranges(
num_event_ranges=num_event_ranges * queue_factor,
job=self.get_job())
for event_range in ret:
self.__event_ranges.append(event_range)
ret = []
for _ in range(num_event_ranges):
if len(self.__event_ranges) > 0:
event_range = self.__event_ranges.pop(0)
ret.append(event_range)
logger.info("Received event ranges(num:%s): %s" % (len(ret), ret))
return ret
def update_events(self, messages):
logger.info("Updating event ranges: %s" % messages)
ret = self.communication_manager.update_events(messages)
logger.info("Updated event ranges status: %s" % ret)
return ret
def update_jobs(self, jobs):
logger.info("Updating jobs: %s" % jobs)
ret = self.communication_manager.update_jobs(jobs)
logger.info("Updated jobs status: %s" % ret)
return ret
def run(self):
"""
Main run process
"""
raise NotImplementedError()
def test_communicator_manager(self):
"""
Make sure that es communicator manager thread works as expected.
"""
communicator_manager = None
try:
args = {
'workflow': 'eventservice_hpc',
'queue': 'BNL_CLOUD_MCORE',
'site': 'BNL_CLOUD_MCORE',
'port': 25443,
'url': 'https://aipanda007.cern.ch',
'job_label': 'ptest',
'pilot_user': '******',
'node': socket.getfqdn(),
'mem': 16000,
'disk_space': 160000,
'working_group': '',
'cpu': 2601.0,
'info': None
}
communicator_manager = CommunicationManager()
communicator_manager.start()
self.assertTrue(communicator_manager.is_alive())
jobs = communicator_manager.get_jobs(njobs=2, args=args)
self.assertEqual(len(jobs), 2)
jobs = communicator_manager.get_jobs(njobs=1, args=args)
self.assertEqual(len(jobs), 1)
job_list = []
for job in jobs:
job_data = {
'node': socket.getfqdn(),
'pilotErrorCode': 0,
'startTime': time.time(),
'jobMetrics': 'coreCount=8',
'schedulerID': 'unknown',
'timestamp': time_stamp(),
'exeErrorCode': 0,
'pilotID': 'unknown|PR|2.0.0 (80)',
'transExitCode': 0,
'pilotErrorDiag': '',
'exeErrorDiag': ''
}
job_data['jobId'] = job['PandaID']
job_data['siteName'] = 'BNL_CLOUD_MCORE'
job_data['state'] = 'running'
job_data['attemptNr'] = job['attemptNr'] + 1
job_list.append(job_data)
status = communicator_manager.update_jobs(jobs=job_list)
self.assertEqual(status[0], True)
events = communicator_manager.get_event_ranges(num_event_ranges=1,
job=jobs[0])
self.assertEqual(len(events), 1)
for event in events:
event_range_status = {
"errorCode": 1220,
"eventRangeID": event['eventRangeID'],
"eventStatus": 'failed'
}
event_range_message = {
'version': 0,
'eventRanges': json.dumps(event_range_status)
}
res = communicator_manager.update_events(
update_events=event_range_message)
self.assertEqual(res['StatusCode'], 0)
events = communicator_manager.get_event_ranges(num_event_ranges=2,
job=jobs[0])
self.assertEqual(len(events), 2)
update_events = []
for event in events:
event_range = {
"eventRangeID": event['eventRangeID'],
"eventStatus": 'finished'
}
update_events.append(event_range)
event_range_status = [{
"zipFile": {
"numEvents": len(update_events),
"objstoreID": 1318,
"adler32": '000000',
"lfn": 'test_file',
"fsize": 100,
"pathConvention": 1000
},
"eventRanges": update_events
}]
event_range_message = {
'version': 1,
'eventRanges': json.dumps(event_range_status)
}
res = communicator_manager.update_events(
update_events=event_range_message)
self.assertEqual(res['StatusCode'], 0)
communicator_manager.stop()
time.sleep(2)
self.assertFalse(communicator_manager.is_alive())
except Exception as ex:
if communicator_manager:
communicator_manager.stop()
raise ex
def setUpClass(cls):
try:
args = {
'workflow': 'eventservice_hpc',
'queue': 'BNL_CLOUD_MCORE',
'site': 'BNL_CLOUD_MCORE',
'port': 25443,
'url': 'https://aipanda007.cern.ch',
'job_label': 'ptest',
'pilot_user': '******',
'node': socket.getfqdn(),
'mem': 16000,
'disk_space': 160000,
'working_group': '',
'cpu': 2601.0,
'info': None
}
communicator_manager = CommunicationManager()
cls._communicator_manager = communicator_manager
communicator_manager.start()
jobs = communicator_manager.get_jobs(njobs=1, args=args)
job = create_job(jobs[0], 'BNL_CLOUD_MCORE')
job.workdir = '/tmp/test_esworkexecutor'
job.corecount = 1
if not os.path.exists(job.workdir):
os.makedirs(job.workdir)
job_data = {}
job_data['jobId'] = job['PandaID']
job_data['siteName'] = 'BNL_CLOUD_MCORE'
job_data['state'] = 'starting'
job_data['attemptNr'] = job['attemptNr'] + 1
job_data['node'] = 'pilot2_test'
job_data['schedulerID'] = 'pilot2_test'
job_data['coreCount'] = 1
status = communicator_manager.update_jobs(jobs=[job_data])
job_data['state'] = 'running'
status = communicator_manager.update_jobs(jobs=[job_data])
communicator_manager.stop()
# download input files
client = StageInESClient(job.infosys, logger=logger)
kwargs = dict(workdir=job.workdir,
cwd=job.workdir,
usecontainer=False,
job=job)
client.prepare_sources(job.indata)
client.transfer(job.indata, activity='pr', **kwargs)
# get the payload command from the user specific code
pilot_user = os.environ.get('PILOT_USER', 'atlas').lower()
user = __import__('pilot.user.%s.common' % pilot_user, globals(),
locals(), [pilot_user], 0) # Python 2/3
cmd = user.get_payload_command(job)
logger.info("payload execution command: %s" % cmd)
payload = {
'executable': cmd,
'workdir': job.workdir,
'output_file': 'pilot_test_%s_stdout.txt' % job['PandaID'],
'error_file': 'pilot_test_%s_stderr.txt' % job['PandaID'],
'job': job
}
cls._payload = payload
except Exception as ex:
if cls._communicator_manager:
cls._communicator_manager.stop()
raise ex