def setUp(self):
super().setUp()
workflow_0 = Workflow(id=0, name='test-workflow-0', project_id=0)
workflow_1 = Workflow(id=1, name='test-workflow-1', project_id=0)
db.session.add_all([workflow_0, workflow_1])
config = workflow_definition_pb2.JobDefinition(
name='test-job').SerializeToString()
job_0 = Job(id=0,
name='raw_data_0',
job_type=JobType.RAW_DATA,
state=JobState.STARTED,
workflow_id=0,
project_id=0,
config=config)
job_1 = Job(id=1,
name='raw_data_1',
job_type=JobType.RAW_DATA,
state=JobState.COMPLETED,
workflow_id=0,
project_id=0,
config=config)
job_2 = Job(id=2,
name='data_join_0',
job_type=JobType.DATA_JOIN,
state=JobState.WAITING,
workflow_id=0,
project_id=0,
config=config)
job_3 = Job(id=3,
name='data_join_1',
job_type=JobType.DATA_JOIN,
state=JobState.COMPLETED,
workflow_id=1,
project_id=0,
config=config)
job_4 = Job(id=4,
name='train_job_0',
job_type=JobType.NN_MODEL_TRANINING,
state=JobState.WAITING,
workflow_id=1,
project_id=0,
config=config)
db.session.add_all([job_0, job_1, job_2, job_3, job_4])
job_dep_0 = JobDependency(src_job_id=job_0.id,
dst_job_id=job_2.id,
dep_index=0)
job_dep_1 = JobDependency(src_job_id=job_1.id,
dst_job_id=job_2.id,
dep_index=1)
job_dep_2 = JobDependency(src_job_id=job_3.id,
dst_job_id=job_4.id,
dep_index=0)
db.session.add_all([job_dep_0, job_dep_1, job_dep_2])
db.session.commit()
def _setup_jobs(self):
if self.forked_from is not None:
trunk = Workflow.query.get(self.forked_from)
assert trunk is not None, \
'Source workflow %d not found' % self.forked_from
trunk_job_defs = trunk.get_config().job_definitions
trunk_name2index = {
job.name: i
for i, job in enumerate(trunk_job_defs)
}
job_defs = self.get_config().job_definitions
flags = self.get_create_job_flags()
assert len(job_defs) == len(flags), \
'Number of job defs does not match number of create_job_flags ' \
'%d vs %d'%(len(job_defs), len(flags))
jobs = []
for i, (job_def, flag) in enumerate(zip(job_defs, flags)):
if flag == common_pb2.CreateJobFlag.REUSE:
assert job_def.name in trunk_name2index, \
f'Job {job_def.name} not found in base workflow'
j = trunk.get_job_ids()[trunk_name2index[job_def.name]]
job = Job.query.get(j)
assert job is not None, \
'Job %d not found' % j
# TODO: check forked jobs does not depend on non-forked jobs
else:
job = Job(
name=f'{self.uuid}-{job_def.name}',
job_type=JobType(job_def.job_type),
config=job_def.SerializeToString(),
workflow_id=self.id,
project_id=self.project_id,
state=JobState.NEW,
is_disabled=(flag == common_pb2.CreateJobFlag.DISABLED))
db.session.add(job)
jobs.append(job)
db.session.flush()
name2index = {job.name: i for i, job in enumerate(job_defs)}
for i, (job, flag) in enumerate(zip(jobs, flags)):
if flag == common_pb2.CreateJobFlag.REUSE:
continue
for j, dep_def in enumerate(job.get_config().dependencies):
dep = JobDependency(
src_job_id=jobs[name2index[dep_def.source]].id,
dst_job_id=job.id,
dep_index=j)
db.session.add(dep)
self.set_job_ids([job.id for job in jobs])
if Features.FEATURE_MODEL_WORKFLOW_HOOK:
for job in jobs:
ModelService(db.session).workflow_hook(job)
def _setup_jobs(self):
if self.forked_from is not None:
trunk = Workflow.query.get(self.forked_from)
assert trunk is not None, \
'Source workflow %d not found'%self.forked_from
trunk_job_defs = trunk.get_config().job_definitions
trunk_name2index = {
job.name: i
for i, job in enumerate(trunk_job_defs)
}
else:
assert not self.get_reuse_job_names()
job_defs = self.get_config().job_definitions
jobs = []
reuse_jobs = set(self.get_reuse_job_names())
for i, job_def in enumerate(job_defs):
if job_def.name in reuse_jobs:
assert job_def.name in trunk_name2index, \
"Job %s not found in base workflow"%job_def.name
j = trunk.get_job_ids()[trunk_name2index[job_def.name]]
job = Job.query.get(j)
assert job is not None, \
'Job %d not found'%j
# TODO: check forked jobs does not depend on non-forked jobs
else:
job = Job(name=f'{self.name}-{job_def.name}',
job_type=JobType(job_def.type),
config=job_def.SerializeToString(),
workflow_id=self.id,
project_id=self.project_id,
state=JobState.STOPPED)
job.set_yaml_template(job_def.yaml_template)
db.session.add(job)
jobs.append(job)
db.session.commit()
name2index = {job.name: i for i, job in enumerate(job_defs)}
for i, job in enumerate(jobs):
if job.name in reuse_jobs:
continue
for j, dep_def in enumerate(job.get_config().dependencies):
dep = JobDependency(
src_job_id=jobs[name2index[dep_def.source]].id,
dst_job_id=job.id,
dep_index=j)
db.session.add(dep)
self.set_job_ids([job.id for job in jobs])
db.session.commit()