def k8s_events_handle_build_job_statuses(self, payload):
"""Project Plugin jobs statuses"""
details = payload['details']
app = details['labels']['app']
job_uuid = details['labels']['job_uuid']
job_name = details['labels']['job_name']
project_name = details['labels'].get('project_name')
logger.debug('handling events status for build jon %s %s', job_name, app)
try:
build_job = BuildJob.objects.get(uuid=job_uuid)
except BuildJob.DoesNotExist:
logger.info('Build job `%s` does not exist', job_name)
return
try:
build_job.project
except Project.DoesNotExist:
logger.debug('`%s` does not exist anymore', project_name)
# Set the new status
try:
set_node_scheduling(build_job, details['node_name'])
build_job.set_status(status=payload['status'],
message=payload['message'],
traceback=payload.get('traceback'),
details=details)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_events_reconcile_job_statuses(job_id, status, created_at) -> None:
try:
job = Job.objects.get(id=job_id)
except ExperimentJob.DoesNotExist:
logger.debug('Job `%s` does not exist', job_id)
return
if job.is_done:
return
job.set_status(status=status,
message='Status was reconciled.',
created_at=created_at)
def k8s_events_reconcile_plugin_job_statuses(job_id, app, status,
created_at) -> None:
job = get_plugin_job(app=app, job_uuid=job_id)
if not job:
logger.debug('Job `%s` does not exist', job_id)
return
if job.is_done:
return
job.set_status(status=status,
message='Status was reconciled.',
created_at=created_at)
def k8s_events_handle_experiment_job_statuses(self, payload):
"""Experiment jobs statuses"""
details = payload['details']
job_uuid = details['labels']['job_uuid']
logger.debug('handling events status for job_uuid: %s, status: %s',
job_uuid, payload['status'])
try:
job = ExperimentJob.objects.get(uuid=job_uuid)
except ExperimentJob.DoesNotExist:
logger.debug('Job uuid`%s` does not exist', job_uuid)
return
try:
job.experiment
except Experiment.DoesNotExist:
logger.debug('Experiment for job `%s` does not exist anymore', job_uuid)
return
if job.last_status is None and self.request.retries < 2:
self.retry(countdown=1)
# Set the new status
try:
set_node_scheduling(job, details['node_name'])
job.set_status(status=payload['status'],
message=payload['message'],
created_at=payload.get('created_at'),
traceback=payload.get('traceback'),
details=details)
logger.debug('status %s is set for job %s %s', payload['status'], job_uuid, job.id)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
logger.info('Retry job status %s handling %s', payload['status'], job_uuid)
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_events_handle_job_statuses(self: 'workers.app.task',
payload: Dict) -> None:
"""Project jobs statuses"""
details = payload['details']
job_uuid = details['labels']['job_uuid']
job_name = details['labels']['job_name']
project_name = details['labels'].get('project_name')
logger.debug('handling events status for job %s', job_name)
try:
job = Job.objects.get(uuid=job_uuid)
except Job.DoesNotExist:
logger.debug('Job `%s` does not exist', job_name)
return
try:
job.project
except Project.DoesNotExist:
logger.debug('Project for job `%s` does not exist', project_name)
return
# Set the new status
try:
RedisStatuses.set_status(job_uuid, payload['status'])
set_node_scheduling(job, details['node_name'])
job.set_status(status=payload['status'],
message=payload['message'],
traceback=payload.get('traceback'),
details=details)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_events_handle_build_job_statuses(self: 'workers.app.task',
payload: Dict) -> None:
"""Project Plugin jobs statuses"""
details = payload['details']
app = details['labels']['app']
job_uuid = details['labels']['job_uuid']
job_name = details['labels']['job_name']
restart_count = payload.get('restart_count', 0)
project_name = details['labels'].get('project_name')
logger.debug('handling events status for build jon %s %s', job_name, app)
try:
build_job = BuildJob.objects.get(uuid=job_uuid)
except BuildJob.DoesNotExist:
logger.info('Build job `%s` does not exist', job_name)
return
try:
build_job.project
except Project.DoesNotExist:
logger.debug('`%s` does not exist anymore', project_name)
max_restarts = build_job.max_restarts or conf.get(MAX_RESTARTS_BUILD_JOBS)
if JobLifeCycle.failed(payload['status']) and restart_count < max_restarts:
return
# Set the new status
try:
RedisStatuses.set_status(job_uuid, payload['status'])
set_node_scheduling(build_job, details['node_name'])
build_job.set_status(status=payload['status'],
message=payload['message'],
traceback=payload.get('traceback'),
details=details)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_events_handle_experiment_job_statuses(self: 'workers.app.task',
payload: Dict) -> None:
"""Experiment jobs statuses"""
details = payload['details']
job_uuid = details['labels']['job_uuid']
restart_count = payload.get('restart_count', 0)
logger.debug('handling events status for job_uuid: %s, status: %s',
job_uuid, payload['status'])
try:
job = ExperimentJob.objects.get(uuid=job_uuid)
except ExperimentJob.DoesNotExist:
logger.debug('Job uuid`%s` does not exist', job_uuid)
return
try:
experiment = job.experiment
except Experiment.DoesNotExist:
logger.debug('Experiment for job `%s` does not exist anymore',
job_uuid)
return
if job.last_status is None and self.request.retries < 2:
self.retry(countdown=1)
max_restarts = experiment.max_restarts or conf.get(
MAX_RESTARTS_EXPERIMENTS)
if JobLifeCycle.failed(payload['status']) and restart_count < max_restarts:
return
# Set the new status
try:
RedisStatuses.set_status(job_uuid, payload['status'])
set_node_scheduling(job, details['node_name'])
job.set_status(status=payload['status'],
message=payload['message'],
created_at=payload.get('created_at'),
traceback=payload.get('traceback'),
details=details)
logger.debug('status %s is set for job %s %s', payload['status'],
job_uuid, job.id)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
logger.info('Retry job status %s handling %s', payload['status'],
job_uuid)
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_events_handle_plugin_job_statuses(self: 'workers.app.task',
payload: Dict) -> None:
"""Project Plugin jobs statuses"""
details = payload['details']
app = details['labels']['app']
job_uuid = details['labels']['job_uuid']
job_name = details['labels']['job_name']
project_name = details['labels'].get('project_name')
logger.debug('handling events status for job %s %s', job_name, app)
try:
if app == conf.get(APP_LABELS_TENSORBOARD):
job = TensorboardJob.objects.get(uuid=job_uuid)
elif app == conf.get(APP_LABELS_NOTEBOOK):
job = NotebookJob.objects.get(uuid=job_uuid)
else:
logger.info('Plugin job `%s` does not exist', app)
return
except (NotebookJob.DoesNotExist, TensorboardJob.DoesNotExist):
logger.debug('`%s - %s` does not exist', app, job_name)
return
try:
job.project
except Project.DoesNotExist:
logger.debug('`%s` does not exist anymore', project_name)
# Set the new status
try:
RedisStatuses.set_status(job_uuid, payload['status'])
set_node_scheduling(job, details['node_name'])
job.set_status(status=payload['status'],
message=payload['message'],
traceback=payload.get('traceback'),
details=details)
except IntegrityError:
# Due to concurrency this could happen, we just retry it
self.retry(countdown=Intervals.EXPERIMENTS_SCHEDULER)
def k8s_handle_events_namespace(cluster_id: int, payload: Dict) -> None:
logger.debug('handling events namespace for cluster: %s', cluster_id)
try:
ClusterEvent.objects.create(cluster_id=cluster_id, **payload)
except OperationalError:
pass
