def function_invoker(event):
if __version__ != event['cloudbutton_version']:
raise Exception("WRONGVERSION", "PyWren version mismatch",
__version__, event['cloudbutton_version'])
if event['log_level']:
cloud_logging_config(event['log_level'])
log_level = logging.getLevelName(logger.getEffectiveLevel())
custom_env = {'CLOUDBUTTON_FUNCTION': 'True',
'PYTHONUNBUFFERED': 'True',
'CLOUDBUTTON_LOGLEVEL': log_level}
os.environ.update(custom_env)
config = event['config']
num_invokers = event['invokers']
invoker = FunctionInvoker(config, num_invokers, log_level)
invoker.run(event['job_description'])
def __init__(self, jr_config, jobrunner_conn, internal_storage):
self.jr_config = jr_config
self.jobrunner_conn = jobrunner_conn
self.internal_storage = internal_storage
log_level = self.jr_config['log_level']
cloud_logging_config(log_level)
self.cloudbutton_config = self.jr_config['cloudbutton_config']
self.call_id = self.jr_config['call_id']
self.job_id = self.jr_config['job_id']
self.executor_id = self.jr_config['executor_id']
self.func_key = self.jr_config['func_key']
self.data_key = self.jr_config['data_key']
self.data_byte_range = self.jr_config['data_byte_range']
self.output_key = self.jr_config['output_key']
self.stats = stats(self.jr_config['stats_filename'])
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import logging
import os
from cloudbutton.config import cloud_logging_config
from cloudbutton.engine.agent.handler import function_handler
cloud_logging_config(logging.INFO)
logger = logging.getLogger('__main__')
def main(event, context):
logger.info("Starting AWS Lambda Function execution")
os.environ['__OW_ACTIVATION_ID'] = context.aws_request_id
os.environ['__PW_ACTIVATION_ID'] = context.aws_request_id
function_handler(event)
return {"Execution": "Finished"}
def function_handler(event):
start_tstamp = time.time()
log_level = event['log_level']
cloud_logging_config(log_level)
logger.debug("Action handler started")
extra_env = event.get('extra_env', {})
os.environ.update(extra_env)
os.environ.update({'CLOUDBUTTON_FUNCTION': 'True',
'PYTHONUNBUFFERED': 'True'})
config = event['config']
call_id = event['call_id']
job_id = event['job_id']
executor_id = event['executor_id']
exec_id = "{}/{}/{}".format(executor_id, job_id, call_id)
logger.info("Execution-ID: {}".format(exec_id))
runtime_name = event['runtime_name']
runtime_memory = event['runtime_memory']
execution_timeout = event['execution_timeout']
logger.debug("Runtime name: {}".format(runtime_name))
logger.debug("Runtime memory: {}MB".format(runtime_memory))
logger.debug("Function timeout: {}s".format(execution_timeout))
func_key = event['func_key']
data_key = event['data_key']
data_byte_range = event['data_byte_range']
storage_config = extract_storage_config(config)
internal_storage = InternalStorage(storage_config)
call_status = CallStatus(config, internal_storage)
call_status.response['host_submit_tstamp'] = event['host_submit_tstamp']
call_status.response['start_tstamp'] = start_tstamp
context_dict = {
'cloudbutton_version': os.environ.get("CLOUDBUTTON_VERSION"),
'call_id': call_id,
'job_id': job_id,
'executor_id': executor_id,
'activation_id': os.environ.get('__PW_ACTIVATION_ID')
}
call_status.response.update(context_dict)
show_memory_peak = strtobool(os.environ.get('SHOW_MEMORY_PEAK', 'False'))
try:
if version.__version__ != event['cloudbutton_version']:
msg = ("Cloudbutton version mismatch. Host version: {} - Runtime version: {}"
.format(event['cloudbutton_version'], version.__version__))
raise RuntimeError('HANDLER', msg)
# send init status event
call_status.send('__init__')
# call_status.response['free_disk_bytes'] = free_disk_space("/tmp")
custom_env = {'CLOUDBUTTON_CONFIG': json.dumps(config),
'CLOUDBUTTON_EXECUTION_ID': exec_id,
'PYTHONPATH': "{}:{}".format(os.getcwd(), LIBS_PATH)}
os.environ.update(custom_env)
jobrunner_stats_dir = os.path.join(STORAGE_FOLDER,
storage_config['bucket'],
JOBS_PREFIX, executor_id,
job_id, call_id)
os.makedirs(jobrunner_stats_dir, exist_ok=True)
jobrunner_stats_filename = os.path.join(jobrunner_stats_dir, 'jobrunner.stats.txt')
jobrunner_config = {'cloudbutton_config': config,
'call_id': call_id,
'job_id': job_id,
'executor_id': executor_id,
'func_key': func_key,
'data_key': data_key,
'log_level': log_level,
'data_byte_range': data_byte_range,
'output_key': create_output_key(JOBS_PREFIX, executor_id, job_id, call_id),
'stats_filename': jobrunner_stats_filename}
if show_memory_peak:
mm_handler_conn, mm_conn = Pipe()
memory_monitor = Thread(target=memory_monitor_worker, args=(mm_conn, ))
memory_monitor.start()
handler_conn, jobrunner_conn = Pipe()
jobrunner = JobRunner(jobrunner_config, jobrunner_conn, internal_storage)
logger.debug('Starting JobRunner process')
local_execution = strtobool(os.environ.get('__PW_LOCAL_EXECUTION', 'False'))
jrp = Thread(target=jobrunner.run) if local_execution else Process(target=jobrunner.run)
jrp.start()
jrp.join(execution_timeout)
logger.debug('JobRunner process finished')
if jrp.is_alive():
# If process is still alive after jr.join(job_max_runtime), kill it
try:
jrp.terminate()
except Exception:
# thread does not have terminate method
pass
msg = ('Function exceeded maximum time of {} seconds and was '
'killed'.format(execution_timeout))
raise TimeoutError('HANDLER', msg)
if show_memory_peak:
mm_handler_conn.send('STOP')
memory_monitor.join()
peak_memory_usage = int(mm_handler_conn.recv())
logger.info("Peak memory usage: {}".format(sizeof_fmt(peak_memory_usage)))
call_status.response['peak_memory_usage'] = peak_memory_usage
if not handler_conn.poll():
logger.error('No completion message received from JobRunner process')
logger.debug('Assuming memory overflow...')
# Only 1 message is returned by jobrunner when it finishes.
# If no message, this means that the jobrunner process was killed.
# 99% of times the jobrunner is killed due an OOM, so we assume here an OOM.
msg = 'Function exceeded maximum memory and was killed'
raise MemoryError('HANDLER', msg)
if os.path.exists(jobrunner_stats_filename):
with open(jobrunner_stats_filename, 'r') as fid:
for l in fid.readlines():
key, value = l.strip().split(" ", 1)
try:
call_status.response[key] = float(value)
except Exception:
call_status.response[key] = value
if key in ['exception', 'exc_pickle_fail', 'result', 'new_futures']:
call_status.response[key] = eval(value)
except Exception:
# internal runtime exceptions
print('----------------------- EXCEPTION !-----------------------', flush=True)
traceback.print_exc(file=sys.stdout)
print('----------------------------------------------------------', flush=True)
call_status.response['exception'] = True
pickled_exc = pickle.dumps(sys.exc_info())
pickle.loads(pickled_exc) # this is just to make sure they can be unpickled
call_status.response['exc_info'] = str(pickled_exc)
finally:
call_status.response['end_tstamp'] = time.time()
call_status.send('__end__')
for key in extra_env:
os.environ.pop(key)
logger.info("Finished")