def create_job_exe_model(self, framework_id, when):
"""Creates and returns a scheduled job execution model
:param framework_id: The scheduling framework ID
:type framework_id: string
:param when: The start time
:type when: :class:`datetime.datetime`
:returns: The job execution model
:rtype: :class:`job.models.JobExecution`
"""
job_exe = JobExecution()
job_exe.job_id = self._queue.job_id
job_exe.job_type_id = self._queue.job_type_id
job_exe.recipe_id = self._queue.recipe_id
job_exe.batch_id = self._queue.batch_id
job_exe.exe_num = self._queue.exe_num
job_exe.timeout = self._queue.timeout
job_exe.input_file_size = self._queue.input_file_size
job_exe.configuration = self.configuration.get_dict()
job_exe.queued = self._queue.queued
if self.is_canceled:
job_exe.node_id = None
job_exe.resources = NodeResources().get_json().get_dict()
job_exe.started = None
else:
job_exe.node_id = self._scheduled_node_id
job_exe.resources = self._scheduled_resources.get_json().get_dict()
job_exe.started = when
job_exe.set_cluster_id(framework_id, self._queue.job_id,
self._queue.exe_num)
return job_exe
def create_running_job_exe(agent_id='agent_1', job_type=None, job=None, node=None, timeout=None, input_file_size=10.0,
queued=None, started=None, resources=None, priority=None, num_exes=1):
"""Creates a running job execution for unit testing
:returns: The running job execution
:rtype: :class:`job.execution.job_exe.RunningJobExecution`
"""
when = timezone.now()
if not job:
job = create_job(job_type=job_type, status='RUNNING', input_file_size=input_file_size, num_exes=num_exes)
job_type = job.job_type
# Configuration that occurs at queue time
input_files = {}
input_file_ids = job.get_job_data().get_input_file_ids()
if input_file_ids:
for input_file in ScaleFile.objects.get_files_for_queued_jobs(input_file_ids):
input_files[input_file.id] = input_file
exe_config = QueuedExecutionConfigurator(input_files).configure_queued_job(job)
job_exe = JobExecution()
job_exe.set_cluster_id('1234', job.id, job.num_exes)
job_exe.job = job
job_exe.job_type = job_type
job_exe.exe_num = job.num_exes
if not node:
node = node_utils.create_node()
job_exe.node = node
if not timeout:
timeout = job.timeout
job_exe.timeout = timeout
job_exe.input_file_size = input_file_size
if not resources:
resources = job.get_resources()
job_exe.resources = resources.get_json().get_dict()
job_exe.configuration = exe_config.get_dict()
if not queued:
queued = when
job_exe.queued = queued
if not started:
started = when + datetime.timedelta(seconds=1)
job_exe.started = started
job_exe.save()
if not priority:
priority = job.priority
# Configuration that occurs at schedule time
workspaces = {}
for workspace in Workspace.objects.all():
workspaces[workspace.name] = workspace
secret_config = ScheduledExecutionConfigurator(workspaces).configure_scheduled_job(job_exe, job_type,
job_type.get_job_interface(),'INFO')
return RunningJobExecution(agent_id, job_exe, job_type, secret_config, priority)
def create_job_exe_model(self, framework_id, when):
"""Creates and returns a scheduled job execution model
:param framework_id: The scheduling framework ID
:type framework_id: string
:param when: The start time
:type when: :class:`datetime.datetime`
:returns: The job execution model
:rtype: :class:`job.models.JobExecution`
"""
job_exe = JobExecution()
job_exe.job_id = self._queue.job_id
job_exe.job_type_id = self._queue.job_type_id
job_exe.recipe_id = self._queue.recipe_id
job_exe.batch_id = self._queue.batch_id
job_exe.exe_num = self._queue.exe_num
job_exe.timeout = self._queue.timeout
job_exe.docker_image = self._queue.docker_image
job_exe.input_file_size = self._queue.input_file_size
job_exe.configuration = self.configuration.get_dict()
job_exe.queued = self._queue.queued
if self.is_canceled:
job_exe.node_id = None
job_exe.resources = NodeResources().get_json().get_dict()
job_exe.started = None
else:
job_exe.node_id = self._scheduled_node_id
job_exe.resources = self._scheduled_resources.get_json().get_dict()
job_exe.started = when
job_exe.set_cluster_id(framework_id, self._queue.job_id, self._queue.exe_num)
if self.required_resources.gpus > 0:
if not GPUManager.assign_gpus_for_job(job_exe.node_id,job_exe.job_id, self.required_resources.gpus):
logger.error("Job %s was unable to assign %s reserved GPUs on node %s. Note: this is not supposed to be able to happen. something has gone horribly wrong.", job_exe.job_id, self.required_resources.gpus, job_exe.node_id)
return job_exe
def create_job_exe(job_type=None,
job=None,
exe_num=None,
node=None,
timeout=None,
input_file_size=10.0,
queued=None,
started=None,
status='RUNNING',
error=None,
ended=None,
output=None,
task_results=None):
"""Creates a job_exe model for unit testing, may also create job_exe_end and job_exe_output models depending on
status
:returns: The job_exe model
:rtype: :class:`job.execution.job_exe.RunningJobExecution`
"""
when = timezone.now()
if not job:
job = create_job(job_type=job_type,
status=status,
input_file_size=input_file_size)
job_type = job.job_type
job_exe = JobExecution()
job_exe.job = job
job_exe.job_type = job_type
if not exe_num:
exe_num = job.num_exes
job_exe.exe_num = exe_num
job_exe.set_cluster_id('1234', job.id, job_exe.exe_num)
if not node:
node = node_utils.create_node()
job_exe.node = node
if not timeout:
timeout = job.timeout
job_exe.timeout = timeout
job_exe.input_file_size = input_file_size
job_exe.resources = job.get_resources().get_json().get_dict()
job_exe.configuration = ExecutionConfiguration().get_dict()
if not queued:
queued = when
job_exe.queued = queued
if not started:
started = when + datetime.timedelta(seconds=1)
job_exe.started = started
job_exe.save()
if status in ['COMPLETED', 'FAILED', 'CANCELED']:
job_exe_end = JobExecutionEnd()
job_exe_end.job_exe_id = job_exe.id
job_exe_end.job = job_exe.job
job_exe_end.job_type = job_exe.job_type
job_exe_end.exe_num = job_exe.exe_num
if not task_results:
task_results = TaskResults()
job_exe_end.task_results = task_results.get_dict()
job_exe_end.status = status
if status == 'FAILED' and not error:
error = error_test_utils.create_error()
job_exe_end.error = error
job_exe_end.node = node
job_exe_end.queued = queued
job_exe_end.started = started
job_exe_end.seed_started = task_results.get_task_started('main')
job_exe_end.seed_ended = task_results.get_task_ended('main')
if not ended:
ended = started + datetime.timedelta(seconds=1)
job_exe_end.ended = ended
job_exe_end.save()
if status == 'COMPLETED' or output:
job_exe_output = JobExecutionOutput()
job_exe_output.job_exe_id = job_exe.id
job_exe_output.job = job_exe.job
job_exe_output.job_type = job_exe.job_type
job_exe_output.exe_num = job_exe.exe_num
if not output:
output = JobResults()
job_exe_output.output = output.get_dict()
job_exe_output.save()
return job_exe