def _add_active_workflow_tokens(self):
"""Add some active workflow tokens.
The job dependencies form a complete binary tree turned upside down.
I.e., each job has two parents.
"""
self._store = EphemeralStore()
version = 1
for level in range(AnalyzerTestCase._NUM_LEVELS):
jobs_at_level = 2**(AnalyzerTestCase._NUM_LEVELS - level - 1)
for job_index in range(jobs_at_level):
job_name = 'job_%d_%d' % (level, job_index)
event_name = Name(workflow='some_workflow',
instance='123',
job=job_name,
event='some_event')
if level == 0:
inputs = [
Name.WORKFLOW_START_INPUT,
Name.WORKFLOW_START_INPUT + '_prime'
]
event_name.input = Name.WORKFLOW_START_INPUT
else:
inputs = [
'job_%d_%d' % (level - 1, 2 * job_index),
'job_%d_%d' % (level - 1, 2 * job_index + 1)
]
event_name.input = 'job_%d_%d' % (level - 1, 2 * job_index)
if level == AnalyzerTestCase._NUM_LEVELS - 1:
outputs = []
else:
outputs = ['job_%d_%d' % (level + 1, job_index / 2)]
job = ShellJob(name=job_name,
inputs=inputs,
outputs=outputs,
command='some_command')
job.history.append(ExecutionRecord())
name = Name(workflow='some_workflow',
instance='123',
job_state=Name.WAITING_STATE,
job=job_name)
job_token = Token(version=version,
name=name.get_job_token_name(),
priority=10,
data=pickle.dumps(job))
version += 1
event = Event('some_event')
event_token = Token(version=version,
name=event_name.get_event_token_name(),
priority=10,
data=pickle.dumps(event))
self._store.commit_tokens([job_token, event_token])
def _get_output_event_tokens(self, job):
"""Create output event tokens for the owned job token.
Args:
job: The job which output tokens should be generated.
Returns:
A list of event tokens corresponding to the outputs of the owned
job token.
"""
assert self._owned_job_token
job_name = Name.from_job_token_name(self._owned_job_token.name)
output_name = Name()
output_name.workflow = job_name.workflow
output_name.instance = job_name.instance
output_name.input = job_name.job
event_tokens = []
for output in job.outputs:
output_name.job = output
output_name.event = get_unique_name()
event = Event(creator=self._name)
assert job.history
execution_record = job.history[-1]
event.attributes = execution_record.get_event_attributes()
event_tokens.append(
Token(name=output_name.get_event_token_name(),
data=pickle.dumps(event)))
return event_tokens
def _get_output_event_tokens(self, job):
"""Create output event tokens for the owned job token.
Args:
job: The job which output tokens should be generated.
Returns:
A list of event tokens corresponding to the outputs of the owned
job token.
"""
assert self._owned_job_token
job_name = Name.from_job_token_name(self._owned_job_token.name)
output_name = Name()
output_name.workflow = job_name.workflow
output_name.instance = job_name.instance
output_name.input = job_name.job
event_tokens = []
for output in job.outputs:
output_name.job = output
output_name.event = get_unique_name()
event = Event(creator=self._name)
assert job.history
execution_record = job.history[-1]
event.attributes = execution_record.get_event_attributes()
event_tokens.append(Token(name=output_name.get_event_token_name(),
data=pickle.dumps(event)))
return event_tokens
def _add_active_workflow_tokens(self):
"""Add some active workflow tokens.
The job dependencies form a complete binary tree turned upside down.
I.e., each job has two parents.
"""
self._store = EphemeralStore()
version = 1
for level in range(AnalyzerTestCase._NUM_LEVELS):
jobs_at_level = 2 ** (AnalyzerTestCase._NUM_LEVELS - level - 1)
for job_index in range(jobs_at_level):
job_name = 'job_{0:d}_{1:d}'.format(level, job_index)
event_name = Name(workflow='some_workflow',
instance='123',
job=job_name,
event='some_event')
if level == 0:
inputs = [Name.WORKFLOW_START_INPUT,
Name.WORKFLOW_START_INPUT + '_prime']
event_name.input = Name.WORKFLOW_START_INPUT
else:
inputs = ['job_{0:d}_{1:d}'.format(level - 1, 2 * job_index),
'job_{0:d}_{1:d}'.format(level - 1, 2 * job_index + 1)]
event_name.input = 'job_{0:d}_{1:d}'.format(level - 1, 2 * job_index)
if level == AnalyzerTestCase._NUM_LEVELS - 1:
outputs = []
else:
outputs = ['job_{0:d}_{1:d}'.format(level + 1, job_index / 2)]
job = ShellJob(name=job_name,
inputs=inputs,
outputs=outputs,
command='some_command')
job.history.append(ExecutionRecord())
name = Name(workflow='some_workflow', instance='123',
job_state=Name.WAITING_STATE, job=job_name)
job_token = Token(version=version,
name=name.get_job_token_name(),
priority=10,
data=pickle.dumps(job))
version += 1
event = Event('some_event')
event_token = Token(version=version,
name=event_name.get_event_token_name(),
priority=10,
data=pickle.dumps(event))
self._store.commit_tokens([job_token, event_token])
def _post_event_tokens(self):
"""Add some event tokens to the master."""
request = ModifyRequest(updates=[])
name = Name(workflow='some_workflow', instance='12345')
for job_id in range(0, 2):
for input_id in range(0, 2):
for event_id in range(0, 2):
name.job = 'some_job_%d' % job_id
name.input = 'some_input_%d' % input_id
name.event = 'some_event_%d' % event_id
event_token = Token(name=name.get_event_token_name())
request.updates.append(event_token)
client = self._factory.get_client()
client.modify(request)
def get_event_names(self, workflow_name, instance, job, input_name):
"""Return names of events under a workflow instance, job, and input."""
request = GroupRequest()
name = Name()
name.workflow = workflow_name
name.instance = instance
name.job = job
name.input = input_name
request.namePrefix = name.get_input_prefix()
request.groupSuffix = Name.DELIMITER
response = self._client.group(request)
events = []
if response.counts:
for event in response.counts.keys():
name = Name.from_event_token_name(event)
events.append(name.event)
return events
def get_event_names(self, workflow_name, instance, job, input_name):
"""Return names of events under a workflow instance, job, and input."""
request = GroupRequest()
name = Name()
name.workflow = workflow_name
name.instance = instance
name.job = job
name.input = input_name
request.namePrefix = name.get_input_prefix()
request.groupSuffix = Name.DELIMITER
response = self._client.group(request)
events = []
if response.counts:
for event in response.counts.keys():
name = Name.from_event_token_name(event)
events.append(name.event)
return events
def _make_job_runnable(self, job_token):
"""Attempt to make a job runnable.
Query event tokens in job inputs. If a combination of triggering
events exist, remove those events and make the job runnable.
Otherwise, do nothing.
Args:
job_token: The job token to make runnable.
Returns:
True if there were no errors during communication with the master,
otherwise False.
"""
job = pickle.loads(job_token.data)
name = Name.from_job_token_name(job_token.name)
request = QueryRequest(queries=[])
# TODO(pawel): handle jobs with no dependencies
assert job.inputs
for input_name in job.inputs:
prefix = Name()
prefix.workflow = name.workflow
prefix.instance = name.instance
prefix.job = name.job
prefix.input = input_name
query = Query()
query.namePrefix = prefix.get_input_prefix()
query.maxTokens = 1
request.queries.append(query)
try:
response = self._client.query(request)
except TokenMasterException:
# TODO(pawel): add a retry count and fail if a limit is reached.
LOG.exception('error sending request %s', request)
return False
triggering_events = Worker._get_triggering_events(response.tokens)
if triggering_events:
return self._move_job_token_to_runnable(job_token,
triggering_events)
return True
def _make_job_runnable(self, job_token):
"""Attempt to make a job runnable.
Query event tokens in job inputs. If a combination of triggering
events exist, remove those events and make the job runnable.
Otherwise, do nothing.
Args:
job_token: The job token to make runnable.
Returns:
True if there were no errors during communication with the master,
otherwise False.
"""
job = pickle.loads(job_token.data)
name = Name.from_job_token_name(job_token.name)
request = QueryRequest(queries=[])
# TODO(pawel): handle jobs with no dependencies
assert job.inputs
for input_name in job.inputs:
prefix = Name()
prefix.workflow = name.workflow
prefix.instance = name.instance
prefix.job = name.job
prefix.input = input_name
query = Query()
query.namePrefix = prefix.get_input_prefix()
query.maxTokens = 1
request.queries.append(query)
try:
response = self._client.query(request)
except TokenMasterException:
# TODO(pawel): add a retry count and fail if a limit is reached.
LOG.exception('error sending request %s', request)
return False
triggering_events = Worker._get_triggering_events(response.tokens)
if triggering_events:
return self._move_job_token_to_runnable(job_token,
triggering_events)
return True
