def create_new_version_task(cls, job, task, dsl_parser, auto):
# stop old version task
FederatedScheduler.stop_task(job=job,
task=task,
stop_status=TaskStatus.CANCELED)
FederatedScheduler.clean_task(
job=job, task=task, content_type=TaskCleanResourceType.METRICS)
# create new version task
task.f_task_version = task.f_task_version + 1
if auto:
task.f_auto_retries = task.f_auto_retries - 1
task.f_run_pid = None
task.f_run_ip = None
# todo: FederatedScheduler.create_task and JobController.initialize_tasks will create task twice
status_code, response = FederatedScheduler.create_task(job=job,
task=task)
if status_code != FederatedSchedulingStatusCode.SUCCESS:
raise Exception(f"create {task.f_task_id} new version failed")
# create the task holder in db to record information of all participants in the initiator for scheduling
for _role in response:
for _party_id in response[_role]:
if _role == job.f_initiator_role and _party_id == job.f_initiator_party_id:
continue
JobController.initialize_tasks(
job_id=job.f_job_id,
role=_role,
party_id=_party_id,
run_on_this_party=False,
initiator_role=job.f_initiator_role,
initiator_party_id=job.f_initiator_party_id,
job_parameters=RunParameters(
**job.f_runtime_conf_on_party["job_parameters"]),
dsl_parser=dsl_parser,
components=[task.f_component_name],
task_version=task.f_task_version,
auto_retries=task.f_auto_retries)
schedule_logger(job.f_job_id).info(
f"create task {task.f_task_id} new version {task.f_task_version} successfully"
)
def rerun_job(cls, job_id, initiator_role, initiator_party_id,
component_name):
schedule_logger(job_id=job_id).info(
f"try to rerun job {job_id} on initiator {initiator_role} {initiator_party_id}"
)
jobs = JobSaver.query_job(job_id=job_id,
role=initiator_role,
party_id=initiator_party_id)
if jobs:
job = jobs[0]
else:
raise RuntimeError(
f"can not found job {job_id} on initiator {initiator_role} {initiator_party_id}"
)
if component_name != job_utils.job_virtual_component_name():
tasks = JobSaver.query_task(job_id=job_id,
role=initiator_role,
party_id=initiator_party_id,
component_name=component_name)
else:
tasks = JobSaver.query_task(job_id=job_id,
role=initiator_role,
party_id=initiator_party_id)
job_can_rerun = False
dsl_parser = schedule_utils.get_job_dsl_parser(
dsl=job.f_dsl,
runtime_conf=job.f_runtime_conf_on_party,
train_runtime_conf=job.f_train_runtime_conf)
for task in tasks:
if task.f_status in {TaskStatus.WAITING, TaskStatus.SUCCESS}:
if task.f_status == TaskStatus.WAITING:
job_can_rerun = True
schedule_logger(job_id=job_id).info(
f"task {task.f_task_id} {task.f_task_version} on {task.f_role} {task.f_party_id} is {task.f_status}, pass rerun"
)
else:
# stop old version task
FederatedScheduler.stop_task(job=job,
task=task,
stop_status=TaskStatus.CANCELED)
FederatedScheduler.clean_task(job=job,
task=task,
content_type="metrics")
# create new version task
task.f_task_version = task.f_task_version + 1
task.f_run_pid = None
task.f_run_ip = None
FederatedScheduler.create_task(job=job, task=task)
# Save the status information of all participants in the initiator for scheduling
schedule_logger(job_id=job_id).info(
f"create task {task.f_task_id} new version {task.f_task_version}"
)
for _role, _party_ids in job.f_runtime_conf_on_party[
"role"].items():
for _party_id in _party_ids:
if _role == initiator_role and _party_id == initiator_party_id:
continue
JobController.initialize_tasks(
job_id,
_role,
_party_id,
False,
job.f_initiator_role,
job.f_initiator_party_id,
RunParameters(
**
job.f_runtime_conf_on_party["job_parameters"]),
dsl_parser,
component_name=task.f_component_name,
task_version=task.f_task_version)
schedule_logger(job_id=job_id).info(
f"create task {task.f_task_id} new version {task.f_task_version} successfully"
)
job_can_rerun = True
if job_can_rerun:
schedule_logger(
job_id=job_id).info(f"job {job_id} set rerun signal")
status = cls.rerun_signal(job_id=job_id, set_or_reset=True)
if status:
schedule_logger(job_id=job_id).info(
f"job {job_id} set rerun signal successfully")
else:
schedule_logger(job_id=job_id).info(
f"job {job_id} set rerun signal failed")
else:
FederatedScheduler.sync_job_status(job=job)
schedule_logger(
job_id=job_id).info(f"job {job_id} no task to rerun")
def submit(cls, job_data, job_id=None):
if not job_id:
job_id = job_utils.generate_job_id()
schedule_logger(job_id).info('submit job, job_id {}, body {}'.format(
job_id, job_data))
job_dsl = job_data.get('job_dsl', {})
job_runtime_conf = job_data.get('job_runtime_conf', {})
job_utils.check_job_runtime_conf(job_runtime_conf)
authentication_utils.check_constraint(job_runtime_conf, job_dsl)
job_initiator = job_runtime_conf['initiator']
conf_adapter = JobRuntimeConfigAdapter(job_runtime_conf)
common_job_parameters = conf_adapter.get_common_parameters()
if common_job_parameters.job_type != 'predict':
# generate job model info
common_job_parameters.model_id = model_utils.gen_model_id(
job_runtime_conf['role'])
common_job_parameters.model_version = job_id
train_runtime_conf = {}
else:
# check predict job parameters
detect_utils.check_config(common_job_parameters.to_dict(),
['model_id', 'model_version'])
# get inference dsl from pipeline model as job dsl
tracker = Tracker(
job_id=job_id,
role=job_initiator['role'],
party_id=job_initiator['party_id'],
model_id=common_job_parameters.model_id,
model_version=common_job_parameters.model_version)
pipeline_model = tracker.get_output_model('pipeline')
train_runtime_conf = json_loads(
pipeline_model['Pipeline'].train_runtime_conf)
if not model_utils.check_if_deployed(
role=job_initiator['role'],
party_id=job_initiator['party_id'],
model_id=common_job_parameters.model_id,
model_version=common_job_parameters.model_version):
raise Exception(
f"Model {common_job_parameters.model_id} {common_job_parameters.model_version} has not been deployed yet."
)
job_dsl = json_loads(pipeline_model['Pipeline'].inference_dsl)
job = Job()
job.f_job_id = job_id
job.f_dsl = job_dsl
job.f_train_runtime_conf = train_runtime_conf
job.f_roles = job_runtime_conf['role']
job.f_work_mode = common_job_parameters.work_mode
job.f_initiator_role = job_initiator['role']
job.f_initiator_party_id = job_initiator['party_id']
job.f_role = job_initiator['role']
job.f_party_id = job_initiator['party_id']
path_dict = job_utils.save_job_conf(
job_id=job_id,
role=job.f_initiator_role,
job_dsl=job_dsl,
job_runtime_conf=job_runtime_conf,
job_runtime_conf_on_party={},
train_runtime_conf=train_runtime_conf,
pipeline_dsl=None)
if job.f_initiator_party_id not in job_runtime_conf['role'][
job.f_initiator_role]:
schedule_logger(job_id).info("initiator party id error:{}".format(
job.f_initiator_party_id))
raise Exception("initiator party id error {}".format(
job.f_initiator_party_id))
# create common parameters on initiator
JobController.backend_compatibility(
job_parameters=common_job_parameters)
JobController.adapt_job_parameters(
role=job.f_initiator_role,
job_parameters=common_job_parameters,
create_initiator_baseline=True)
job.f_runtime_conf = conf_adapter.update_common_parameters(
common_parameters=common_job_parameters)
dsl_parser = schedule_utils.get_job_dsl_parser(
dsl=job.f_dsl,
runtime_conf=job.f_runtime_conf,
train_runtime_conf=job.f_train_runtime_conf)
# initiator runtime conf as template
job.f_runtime_conf_on_party = job.f_runtime_conf.copy()
job.f_runtime_conf_on_party[
"job_parameters"] = common_job_parameters.to_dict()
if common_job_parameters.work_mode == WorkMode.CLUSTER:
# Save the status information of all participants in the initiator for scheduling
for role, party_ids in job.f_roles.items():
for party_id in party_ids:
if role == job.f_initiator_role and party_id == job.f_initiator_party_id:
continue
JobController.initialize_tasks(job_id, role, party_id,
False, job.f_initiator_role,
job.f_initiator_party_id,
common_job_parameters,
dsl_parser)
status_code, response = FederatedScheduler.create_job(job=job)
if status_code != FederatedSchedulingStatusCode.SUCCESS:
job.f_status = JobStatus.FAILED
job.f_tag = "submit_failed"
FederatedScheduler.sync_job_status(job=job)
raise Exception("create job failed", response)
schedule_logger(job_id).info(
'submit job successfully, job id is {}, model id is {}'.format(
job.f_job_id, common_job_parameters.model_id))
logs_directory = job_utils.get_job_log_directory(job_id)
submit_result = {
"job_id":
job_id,
"model_info": {
"model_id": common_job_parameters.model_id,
"model_version": common_job_parameters.model_version
},
"logs_directory":
logs_directory,
"board_url":
job_utils.get_board_url(job_id, job_initiator['role'],
job_initiator['party_id'])
}
submit_result.update(path_dict)
return submit_result
def submit(cls, submit_job_conf: JobConfigurationBase, job_id: str = None):
if not job_id:
job_id = job_utils.generate_job_id()
submit_result = {"job_id": job_id}
schedule_logger(job_id).info(
f"submit job, body {submit_job_conf.to_dict()}")
try:
dsl = submit_job_conf.dsl
runtime_conf = deepcopy(submit_job_conf.runtime_conf)
job_utils.check_job_runtime_conf(runtime_conf)
authentication_utils.check_constraint(runtime_conf, dsl)
job_initiator = runtime_conf["initiator"]
conf_adapter = JobRuntimeConfigAdapter(runtime_conf)
common_job_parameters = conf_adapter.get_common_parameters()
if common_job_parameters.job_type != "predict":
# generate job model info
conf_version = schedule_utils.get_conf_version(runtime_conf)
if conf_version != 2:
raise Exception(
"only the v2 version runtime conf is supported")
common_job_parameters.model_id = model_utils.gen_model_id(
runtime_conf["role"])
common_job_parameters.model_version = job_id
train_runtime_conf = {}
else:
# check predict job parameters
detect_utils.check_config(common_job_parameters.to_dict(),
["model_id", "model_version"])
# get inference dsl from pipeline model as job dsl
tracker = Tracker(
job_id=job_id,
role=job_initiator["role"],
party_id=job_initiator["party_id"],
model_id=common_job_parameters.model_id,
model_version=common_job_parameters.model_version)
pipeline_model = tracker.get_pipeline_model()
train_runtime_conf = json_loads(
pipeline_model.train_runtime_conf)
if not model_utils.check_if_deployed(
role=job_initiator["role"],
party_id=job_initiator["party_id"],
model_id=common_job_parameters.model_id,
model_version=common_job_parameters.model_version):
raise Exception(
f"Model {common_job_parameters.model_id} {common_job_parameters.model_version} has not been deployed yet."
)
dsl = json_loads(pipeline_model.inference_dsl)
# dsl = ProviderManager.fill_fate_flow_provider(dsl)
job = Job()
job.f_job_id = job_id
job.f_dsl = dsl
job.f_train_runtime_conf = train_runtime_conf
job.f_roles = runtime_conf["role"]
job.f_initiator_role = job_initiator["role"]
job.f_initiator_party_id = job_initiator["party_id"]
job.f_role = job_initiator["role"]
job.f_party_id = job_initiator["party_id"]
path_dict = job_utils.save_job_conf(
job_id=job_id,
role=job.f_initiator_role,
party_id=job.f_initiator_party_id,
dsl=dsl,
runtime_conf=runtime_conf,
runtime_conf_on_party={},
train_runtime_conf=train_runtime_conf,
pipeline_dsl=None)
if job.f_initiator_party_id not in runtime_conf["role"][
job.f_initiator_role]:
msg = f"initiator party id {job.f_initiator_party_id} not in roles {runtime_conf['role']}"
schedule_logger(job_id).info(msg)
raise Exception(msg)
# create common parameters on initiator
JobController.create_common_job_parameters(
job_id=job.f_job_id,
initiator_role=job.f_initiator_role,
common_job_parameters=common_job_parameters)
job.f_runtime_conf = conf_adapter.update_common_parameters(
common_parameters=common_job_parameters)
dsl_parser = schedule_utils.get_job_dsl_parser(
dsl=job.f_dsl,
runtime_conf=job.f_runtime_conf,
train_runtime_conf=job.f_train_runtime_conf)
# initiator runtime conf as template
job.f_runtime_conf_on_party = job.f_runtime_conf.copy()
job.f_runtime_conf_on_party[
"job_parameters"] = common_job_parameters.to_dict()
# inherit job
job.f_inheritance_info = common_job_parameters.inheritance_info
job.f_inheritance_status = JobInheritanceStatus.WAITING if common_job_parameters.inheritance_info else JobInheritanceStatus.PASS
if job.f_inheritance_info:
inheritance_jobs = JobSaver.query_job(
job_id=job.f_inheritance_info.get("job_id"),
role=job_initiator["role"],
party_id=job_initiator["party_id"])
inheritance_tasks = JobSaver.query_task(
job_id=job.f_inheritance_info.get("job_id"),
role=job_initiator["role"],
party_id=job_initiator["party_id"],
only_latest=True)
job_utils.check_job_inheritance_parameters(
job, inheritance_jobs, inheritance_tasks)
status_code, response = FederatedScheduler.create_job(job=job)
if status_code != FederatedSchedulingStatusCode.SUCCESS:
job.f_status = JobStatus.FAILED
job.f_tag = "submit_failed"
FederatedScheduler.sync_job_status(job=job)
raise Exception("create job failed", response)
else:
need_run_components = {}
for role in response:
need_run_components[role] = {}
for party, res in response[role].items():
need_run_components[role][party] = [
name for name, value in response[role][party]
["data"]["components"].items()
if value["need_run"] is True
]
if common_job_parameters.federated_mode == FederatedMode.MULTIPLE:
# create the task holder in db to record information of all participants in the initiator for scheduling
for role, party_ids in job.f_roles.items():
for party_id in party_ids:
if role == job.f_initiator_role and party_id == job.f_initiator_party_id:
continue
if not need_run_components[role][party_id]:
continue
JobController.initialize_tasks(
job_id=job_id,
role=role,
party_id=party_id,
run_on_this_party=False,
initiator_role=job.f_initiator_role,
initiator_party_id=job.f_initiator_party_id,
job_parameters=common_job_parameters,
dsl_parser=dsl_parser,
components=need_run_components[role][party_id])
job.f_status = JobStatus.WAITING
status_code, response = FederatedScheduler.sync_job_status(
job=job)
if status_code != FederatedSchedulingStatusCode.SUCCESS:
raise Exception("set job to waiting status failed")
schedule_logger(job_id).info(
f"submit job successfully, job id is {job.f_job_id}, model id is {common_job_parameters.model_id}"
)
logs_directory = job_utils.get_job_log_directory(job_id)
result = {
"code":
RetCode.SUCCESS,
"message":
"success",
"model_info": {
"model_id": common_job_parameters.model_id,
"model_version": common_job_parameters.model_version
},
"logs_directory":
logs_directory,
"board_url":
job_utils.get_board_url(job_id, job_initiator["role"],
job_initiator["party_id"])
}
warn_parameter = JobRuntimeConfigAdapter(
submit_job_conf.runtime_conf).check_removed_parameter()
if warn_parameter:
result[
"message"] = f"[WARN]{warn_parameter} is removed,it does not take effect!"
submit_result.update(result)
submit_result.update(path_dict)
except Exception as e:
submit_result["code"] = RetCode.OPERATING_ERROR
submit_result["message"] = exception_to_trace_string(e)
schedule_logger(job_id).exception(e)
return submit_result