class __MPIInterfaceImpl:
""" Implementation of the MPIInterface singleton interface """
def __init__(self):
# Reference to inner cluster object (equivalent to parallel_go)
self._cluster = MPIInterfaceCore()
# Direct reference to MPICommandClient for methods not resorting to the inner cluster
self.__command_client = MPICommandClient()
def isClusterRunning(self):
if self.__command_client.get_lifecyle_state() == 1:
return True
else:
return False
def init_cluster(self, clusterfile=None, project=None):
# NOTE: In the MPI framework the clusterfile is processed by mpirun
# So it is not necessary to process and validate clusterfile here
self.start_cluster()
def start_cluster(self):
# TODO: This should set OMP_NUM_THREADS as well
self._cluster.start_cluster()
def stop_cluster(self):
self._cluster.stop_cluster()
def do_and_record(self, cmd, id=None, group='', subMS=''):
jobId = self._cluster.odo(cmd, id)
return jobId
def get_engine_store(self, id):
res = self._cluster.execute('os.getcwd()', id)
return res[0]['ret']
def get_status(self):
command_request_list = self._cluster.get_command_request_list()
command_response_list = self._cluster.get_command_response_list()
server_status = self._cluster.get_server_status()
# Show first jobs running, then the ones holding the queue and finally the ones already processed
status_list = [
'request sent', 'holding queue', 'timeout', 'response received'
]
# Generate job status table
job_status_list = [[
'JOB ID', 'SERVER', 'HOSTNAME', 'QUEUE STATUS', 'COMMAND',
'ELAPSED TIME (s)', 'EXECUTION STATUS'
]]
for status in status_list:
for jobId in command_request_list:
# Generate job status list
if command_request_list[jobId]['status'] == status:
# Get server to access info from the server status
server = command_request_list[jobId]['server']
hostname = server_status[server]['processor']
# Create job status info
job_status = []
job_status.append(str(jobId))
job_status.append(str(server))
job_status.append(hostname)
job_status.append(
command_request_list[jobId]['status'])
job_status.append(
command_request_list[jobId]['command'])
# Add run time
elapsed_time = ''
if status == status_list[1]: # holding queue
elapsed_time = 'N/A'
elif status == status_list[0] or status == status_list[
2]: # request sent / timeout
start_time = server_status[server][
'command_start_time']
if start_time is not None:
elapsed_time = "%.2f" % (time.time() -
start_time)
else:
elapsed_time = 'unset'
elif status == status_list[3]: # 'response received'
start_time = command_response_list[jobId][
'command_start_time']
stop_time = command_response_list[jobId][
'command_stop_time']
if start_time is not None and stop_time is not None:
elapsed_time = "%.2f" % (stop_time -
start_time)
else:
elapsed_time = 'unset'
job_status.append(elapsed_time)
# Add job execution status
execution_status = ''
if status == status_list[0] or status == status_list[
1]: # request sent / holding queue
execution_status = 'pending'
elif status == status_list[2]: # timeout
execution_status = 'timeout'
elif status == status_list[3]: # response received
if command_response_list[jobId]['successful']:
ret = command_response_list[jobId]['ret']
if isinstance(ret, bool):
if ret is True:
execution_status = 'completed - True'
else:
execution_status = 'completed - False'
else:
execution_status = 'completed'
else:
execution_status = 'exception raised'
job_status.append(execution_status)
# Append job status info to list
job_status_list.append(job_status)
# Print job status table
MPIInterface.print_table(job_status_list)
# Return job status table
return job_status_list
def set_log_level(self, log_level):
self.__command_client.set_log_level(log_level)
class __MPIInterfaceImpl:
""" Implementation of the MPIInterface singleton interface """
def __init__(self):
# Reference to inner cluster object (equivalent to parallel_go)
self._cluster = MPIInterfaceCore()
# Direct reference to MPICommandClient for methods not resorting to the inner cluster
self.__command_client = MPICommandClient()
def isClusterRunning(self):
if self.__command_client.get_lifecyle_state() == 1:
return True
else:
return False
def init_cluster(self, clusterfile=None, project=None):
# NOTE: In the MPI framework the clusterfile is processed by mpirun
# So it is not necessary to process and validate clusterfile here
self.start_cluster()
def start_cluster(self):
# TODO: This should set OMP_NUM_THREADS as well
self._cluster.start_cluster()
def stop_cluster(self):
self._cluster.stop_cluster()
def do_and_record(self, cmd, id=None, group='', subMS=''):
jobId = self._cluster.odo(cmd,id)
return jobId
def get_engine_store(self, id):
res = self._cluster.execute('os.getcwd()',id)
return res[0]['ret']
def get_status(self):
command_request_list = self._cluster.get_command_request_list()
command_response_list = self._cluster.get_command_response_list()
server_status = self._cluster.get_server_status()
# Show first jobs running, then the ones holding the queue and finally the ones already processed
status_list = ['request sent','holding queue','timeout','response received']
# Generate job status table
job_status_list = [['JOB ID','SERVER','HOSTNAME','QUEUE STATUS','COMMAND','ELAPSED TIME (s)','EXECUTION STATUS']]
for status in status_list:
for jobId in command_request_list:
# Generate job status list
if command_request_list[jobId]['status'] == status:
# Get server to access info from the server status
server = command_request_list[jobId]['server']
hostname = server_status[server]['processor']
# Create job status info
job_status = []
job_status.append(str(jobId))
job_status.append(str(server))
job_status.append(hostname)
job_status.append(command_request_list[jobId]['status'])
job_status.append(command_request_list[jobId]['command'])
# Add run time
elapsed_time = ''
if status == status_list[1]: # holding queue
elapsed_time = 'N/A'
elif status == status_list[0] or status == status_list[2]: # request sent / timeout
start_time = server_status[server]['command_start_time']
if start_time is not None:
elapsed_time = "%.2f" % (time.time() - start_time)
else:
elapsed_time = 'unset'
elif status == status_list[3]: # 'response received'
start_time = command_response_list[jobId]['command_start_time']
stop_time = command_response_list[jobId]['command_stop_time']
if start_time is not None and stop_time is not None:
elapsed_time = "%.2f" % (stop_time- start_time)
else:
elapsed_time = 'unset'
job_status.append(elapsed_time)
# Add job execution status
execution_status = ''
if status == status_list[0] or status == status_list[1]: # request sent / holding queue
execution_status = 'pending'
elif status == status_list[2]: # timeout
execution_status = 'timeout'
elif status == status_list[3]: # response received
if command_response_list[jobId]['successful']:
ret = command_response_list[jobId]['ret']
if isinstance(ret,bool):
if ret == True:
execution_status = 'completed - True'
else:
execution_status = 'completed - False'
else:
execution_status = 'completed'
else:
execution_status = 'exception raised'
job_status.append(execution_status)
# Append job status info to list
job_status_list.append(job_status)
# Print job status table
MPIInterface.print_table(job_status_list)
# Return job status table
return job_status_list
def set_log_level(self,log_level):
self.__command_client.set_log_level(log_level)
