def test_job_monitoring(self):
job_uuid = uuid.uuid4().hex
assert RedisToStream.is_monitored_job_resources(job_uuid) is False
RedisToStream.monitor_job_resources(job_uuid)
assert RedisToStream.is_monitored_job_resources(job_uuid) is True
RedisToStream.remove_job_resources(job_uuid)
assert RedisToStream.is_monitored_job_resources(job_uuid) is False
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
RedisToStream.monitor_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is True
RedisToStream.remove_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
def test_job_monitoring(self):
job_uuid = uuid.uuid4().hex
assert RedisToStream.is_monitored_job_resources(job_uuid) is False
RedisToStream.monitor_job_resources(job_uuid)
assert RedisToStream.is_monitored_job_resources(job_uuid) is True
RedisToStream.remove_job_resources(job_uuid)
assert RedisToStream.is_monitored_job_resources(job_uuid) is False
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
RedisToStream.monitor_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is True
RedisToStream.remove_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
async def job_logs(request, ws, username, project_name, experiment_sequence,
job_sequence):
project = _get_project(username, project_name)
if not has_project_permissions(request.app.user, project, 'GET'):
exceptions.Forbidden("You don't have access to this project")
experiment = _get_validated_experiment(project, experiment_sequence)
job = _get_job(experiment, job_sequence)
job_uuid = job.uuid.hex
auditor.record(event_type=EXPERIMENT_JOB_LOGS_VIEWED,
instance=job,
actor_id=request.app.user.id)
if not RedisToStream.is_monitored_job_logs(job_uuid=job_uuid):
logger.info('Job uuid `%s` logs is now being monitored', job_uuid)
RedisToStream.monitor_job_logs(job_uuid=job_uuid)
# start consumer
if job_uuid in request.app.job_logs_consumers:
consumer = request.app.job_logs_consumers[job_uuid]
else:
logger.info('Add job log consumer for %s', job_uuid)
consumer = Consumer(
routing_key='{}.{}.{}'.format(RoutingKeys.LOGS_SIDECARS,
experiment.uuid.hex, job_uuid),
queue='{}.{}'.format(CeleryQueues.STREAM_LOGS_SIDECARS, job_uuid))
request.app.job_logs_consumers[job_uuid] = consumer
consumer.run()
# add socket manager
consumer.add_socket(ws)
should_quite = False
num_message_retries = 0
while True:
num_message_retries += 1
for message in consumer.get_messages():
num_message_retries = 0
disconnected_ws = set()
for _ws in consumer.ws:
try:
await _ws.send(message)
except ConnectionClosed:
disconnected_ws.add(_ws)
consumer.remove_sockets(disconnected_ws)
# After trying a couple of time, we must check the status of the experiment
if num_message_retries > MAX_RETRIES:
job.refresh_from_db()
if job.is_done:
logger.info('removing all socket because the job `%s` is done',
job_uuid)
consumer.ws = set([])
else:
num_message_retries -= CHECK_DELAY
# Just to check if connection closed
if ws._connection_lost: # pylint:disable=protected-access
logger.info('Quitting logs socket for job uuid %s', job_uuid)
consumer.remove_sockets({
ws,
})
should_quite = True
if not consumer.ws:
logger.info('Stopping logs monitor for job uuid %s', job_uuid)
RedisToStream.remove_job_logs(job_uuid=job_uuid)
# if job_uuid in request.app.job_logs_consumers:
# consumer = request.app.job_logs_consumers.pop(job_uuid, None)
# if consumer:
# consumer.stop()
should_quite = True
if should_quite:
return
await asyncio.sleep(SOCKET_SLEEP)
def test_monitor_job_logs(self):
job_uuid = uuid.uuid4().hex
RedisToStream.monitor_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is True
RedisToStream.remove_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
async def build_logs(
request, # pylint:disable=too-many-branches
ws,
username,
project_name,
build_id):
job, message = validate_build(request=request,
username=username,
project_name=project_name,
build_id=build_id)
if job is None:
await ws.send(get_error_message(message))
return
job_uuid = job.uuid.hex
auditor.record(event_type=BUILD_JOB_LOGS_VIEWED,
instance=job,
actor_id=request.app.user.id,
actor_name=request.app.user.username)
if not RedisToStream.is_monitored_job_logs(job_uuid=job_uuid):
_logger.info('Job uuid `%s` logs is now being monitored', job_uuid)
RedisToStream.monitor_job_logs(job_uuid=job_uuid)
# start consumer
if job_uuid in request.app.job_logs_consumers:
consumer = request.app.job_logs_consumers[job_uuid]
else:
_logger.info('Add job log consumer for %s', job_uuid)
consumer = Consumer(
routing_key='{}.{}'.format(RoutingKeys.LOGS_SIDECARS_BUILDS,
job_uuid),
queue='{}.{}'.format(CeleryQueues.STREAM_LOGS_SIDECARS, job_uuid))
request.app.job_logs_consumers[job_uuid] = consumer
consumer.run()
def should_disconnect():
if not consumer.ws:
_logger.info('Stopping logs monitor for job uuid %s', job_uuid)
RedisToStream.remove_job_logs(job_uuid=job_uuid)
# if job_uuid in request.app.job_logs_consumers:
# consumer = request.app.job_logs_consumers.pop(job_uuid, None)
# if consumer:
# consumer.stop()
return True
return False
# add socket manager
consumer.add_socket(ws)
should_quite = False
num_message_retries = 0
# Stream phase changes
status = None
while status != JobLifeCycle.RUNNING and not JobLifeCycle.is_done(status):
job.refresh_from_db()
if status != job.last_status:
status = job.last_status
await notify(consumer=consumer, message=get_status_message(status))
if should_disconnect():
return
await asyncio.sleep(SOCKET_SLEEP)
if JobLifeCycle.is_done(status):
await notify(consumer=consumer, message=get_status_message(status))
RedisToStream.remove_job_logs(job_uuid=job_uuid)
return
while True:
num_message_retries += 1
for message in consumer.get_messages():
num_message_retries = 0
await notify(consumer=consumer, message=message)
# After trying a couple of time, we must check the status of the experiment
if num_message_retries > MAX_RETRIES:
job.refresh_from_db()
if job.is_done:
_logger.info(
'removing all socket because the job `%s` is done',
job_uuid)
consumer.ws = set([])
else:
num_message_retries -= CHECK_DELAY
# Just to check if connection closed
if ws._connection_lost: # pylint:disable=protected-access
_logger.info('Quitting logs socket for job uuid %s', job_uuid)
consumer.remove_sockets({
ws,
})
should_quite = True
if should_disconnect():
should_quite = True
if should_quite:
return
await asyncio.sleep(SOCKET_SLEEP)
def test_monitor_job_logs(self):
job_uuid = uuid.uuid4().hex
RedisToStream.monitor_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is True
RedisToStream.remove_job_logs(job_uuid)
assert RedisToStream.is_monitored_job_logs(job_uuid) is False
async def job_logs(request, ws, username, project_name, experiment_sequence, job_sequence):
project = _get_project(username, project_name)
if not has_project_permissions(request.app.user, project, 'GET'):
exceptions.Forbidden("You don't have access to this project")
experiment = _get_validated_experiment(project, experiment_sequence)
job = _get_job(experiment, job_sequence)
job_uuid = job.uuid.hex
auditor.record(event_type=EXPERIMENT_JOB_LOGS_VIEWED,
instance=job,
actor_id=request.app.user.id)
if not RedisToStream.is_monitored_job_logs(job_uuid=job_uuid):
logger.info('Job uuid `%s` logs is now being monitored', job_uuid)
RedisToStream.monitor_job_logs(job_uuid=job_uuid)
# start consumer
if job_uuid in request.app.job_logs_consumers:
consumer = request.app.job_logs_consumers[job_uuid]
else:
logger.info('Add job log consumer for %s', job_uuid)
consumer = Consumer(
routing_key='{}.{}.{}'.format(RoutingKeys.LOGS_SIDECARS,
experiment.uuid.hex,
job_uuid),
queue='{}.{}'.format(CeleryQueues.STREAM_LOGS_SIDECARS, job_uuid))
request.app.job_logs_consumers[job_uuid] = consumer
consumer.run()
# add socket manager
consumer.add_socket(ws)
should_quite = False
num_message_retries = 0
while True:
num_message_retries += 1
for message in consumer.get_messages():
num_message_retries = 0
disconnected_ws = set()
for _ws in consumer.ws:
try:
await _ws.send(message)
except ConnectionClosed:
disconnected_ws.add(_ws)
consumer.remove_sockets(disconnected_ws)
# After trying a couple of time, we must check the status of the experiment
if num_message_retries > MAX_RETRIES:
job.refresh_from_db()
if job.is_done:
logger.info('removing all socket because the job `%s` is done', job_uuid)
consumer.ws = set([])
else:
num_message_retries -= CHECK_DELAY
# Just to check if connection closed
if ws._connection_lost: # pylint:disable=protected-access
logger.info('Quitting logs socket for job uuid %s', job_uuid)
consumer.remove_sockets({ws, })
should_quite = True
if not consumer.ws:
logger.info('Stopping logs monitor for job uuid %s', job_uuid)
RedisToStream.remove_job_logs(job_uuid=job_uuid)
# if job_uuid in request.app.job_logs_consumers:
# consumer = request.app.job_logs_consumers.pop(job_uuid, None)
# if consumer:
# consumer.stop()
should_quite = True
if should_quite:
return
await asyncio.sleep(SOCKET_SLEEP)