def rse_update(once=False,
process=0,
total_processes=1,
thread=0,
threads_per_process=1):
"""
Main loop to check and update the RSE Counters.
"""
logging.info('rse_update: starting')
logging.info('rse_update: started')
while not graceful_stop.is_set():
try:
# Select a bunch of rses for to update for this worker
start = time.time() # NOQA
rse_ids = get_updated_rse_counters(
total_workers=total_processes * threads_per_process - 1,
worker_number=process * threads_per_process + thread)
logging.debug('Index query time %f size=%d' %
(time.time() - start, len(rse_ids)))
# If the list is empty, sent the worker to sleep
if not rse_ids and not once:
logging.info('rse_update[%s/%s] did not get any work' %
(process * threads_per_process + thread,
total_processes * threads_per_process - 1))
time.sleep(10)
else:
for rse_id in rse_ids:
if graceful_stop.is_set():
break
start_time = time.time()
update_rse_counter(rse_id=rse_id)
logging.debug(
'rse_update[%s/%s]: update of rse "%s" took %f' %
(process * threads_per_process + thread,
total_processes * threads_per_process - 1, rse_id,
time.time() - start_time))
except Exception:
logging.error(traceback.format_exc())
if once:
break
logging.info('rse_update: graceful stop requested')
logging.info('rse_update: graceful stop done')
def rse_update(once=False):
"""
Main loop to check and update the RSE Counters.
"""
logging.info('rse_update: starting')
logging.info('rse_update: started')
# Make an initial heartbeat so that all abacus-rse daemons have the correct worker number on the next try
executable = 'abacus-rse'
hostname = socket.gethostname()
pid = os.getpid()
current_thread = threading.current_thread()
live(executable=executable, hostname=hostname, pid=pid, thread=current_thread)
while not graceful_stop.is_set():
try:
# Heartbeat
heartbeat = live(executable=executable, hostname=hostname, pid=pid, thread=current_thread)
# Select a bunch of rses for to update for this worker
start = time.time() # NOQA
rse_ids = get_updated_rse_counters(total_workers=heartbeat['nr_threads'],
worker_number=heartbeat['assign_thread'])
logging.debug('Index query time %f size=%d' % (time.time() - start, len(rse_ids)))
# If the list is empty, sent the worker to sleep
if not rse_ids and not once:
logging.info('rse_update[%s/%s] did not get any work' % (heartbeat['assign_thread'], heartbeat['nr_threads'] - 1))
time.sleep(10)
else:
for rse_id in rse_ids:
if graceful_stop.is_set():
break
start_time = time.time()
update_rse_counter(rse_id=rse_id)
logging.debug('rse_update[%s/%s]: update of rse "%s" took %f' % (heartbeat['assign_thread'], heartbeat['nr_threads'] - 1, rse_id, time.time() - start_time))
except Exception:
logging.error(traceback.format_exc())
if once:
break
logging.info('rse_update: graceful stop requested')
die(executable=executable, hostname=hostname, pid=pid, thread=current_thread)
logging.info('rse_update: graceful stop done')
def run_once(heartbeat_handler, **_kwargs):
worker_number, total_workers, logger = heartbeat_handler.live()
# Select a bunch of rses for to update for this worker
start = time.time() # NOQA
rse_ids = get_updated_rse_counters(total_workers=total_workers,
worker_number=worker_number)
logger(logging.DEBUG, 'Index query time %f size=%d' % (time.time() - start, len(rse_ids)))
# If the list is empty, sent the worker to sleep
if not rse_ids:
logger(logging.INFO, 'did not get any work')
return
for rse_id in rse_ids:
worker_number, total_workers, logger = heartbeat_handler.live()
if graceful_stop.is_set():
break
start_time = time.time()
update_rse_counter(rse_id=rse_id)
logger(logging.DEBUG, 'update of rse "%s" took %f' % (rse_id, time.time() - start_time))