class Running(object):
def __init__(self,
scheduler,
name,
user='',
master=os.getenv('MESOS_MASTER'),
implicit_acknowledge=1,
*args,
**kwargs):
scheduler = SchedulerProxy(scheduler)
framework = FrameworkInfo(name=name, user=user, *args, **kwargs)
self.driver = MesosSchedulerDriver(scheduler, encode(framework),
master, implicit_acknowledge)
def shutdown(signal, frame):
self.driver.stop()
signal.signal(signal.SIGINT, shutdown)
signal.signal(signal.SIGTERM, shutdown)
atexit.register(self.driver.stop)
def run(self):
return self.driver.run()
def start(self):
status = self.driver.start()
assert status == mesos_pb2.DRIVER_RUNNING
return status
def stop(self):
logging.info("Stopping Mesos driver")
self.driver.stop()
logging.info("Joining Mesos driver")
result = self.driver.join()
logging.info("Joined Mesos driver")
if result != mesos_pb2.DRIVER_STOPPED:
raise RuntimeError("Mesos driver failed with %i", result)
def join(self):
return self.driver.join()
def __enter__(self):
self.start()
return self
def __exit__(self, type, value, traceback):
self.stop()
class Running(object):
def __init__(self, scheduler, name, user='', master=os.getenv('MESOS_MASTER'),
implicit_acknowledge=1, *args, **kwargs):
scheduler = SchedulerProxy(scheduler)
framework = FrameworkInfo(name=name, user=user, *args, **kwargs)
self.driver = MesosSchedulerDriver(scheduler, encode(framework),
master, implicit_acknowledge)
def shutdown(signal, frame):
self.driver.stop()
signal.signal(signal.SIGINT, shutdown)
signal.signal(signal.SIGTERM, shutdown)
atexit.register(self.driver.stop)
def run(self):
return self.driver.run()
def start(self):
status = self.driver.start()
assert status == mesos_pb2.DRIVER_RUNNING
return status
def stop(self):
logging.info("Stopping Mesos driver")
self.driver.stop()
logging.info("Joining Mesos driver")
result = self.driver.join()
logging.info("Joined Mesos driver")
if result != mesos_pb2.DRIVER_STOPPED:
raise RuntimeError("Mesos driver failed with %i", result)
def join(self):
return self.driver.join()
def __enter__(self):
self.start()
return self
def __exit__(self, type, value, traceback):
self.stop()
class Running(object):
def __init__(self, scheduler, name, user='', master=os.getenv('MESOS_MASTER'),
implicit_acknowledge=1, *args, **kwargs):
scheduler = SchedulerProxy(scheduler)
framework = FrameworkInfo(name=name, user=user, *args, **kwargs)
self.driver = MesosSchedulerDriver(scheduler, encode(framework),
master, implicit_acknowledge)
def shutdown(signal, frame):
self.stop()
signal.signal(signal.SIGINT, shutdown)
signal.signal(signal.SIGTERM, shutdown)
atexit.register(self.stop)
def run(self):
return self.driver.run()
def start(self):
status = self.driver.start()
assert status == mesos_pb2.DRIVER_RUNNING
return status
def stop(self):
return self.driver.stop()
def join(self):
return self.driver.join()
def __enter__(self):
self.start()
return self
def __exit__(self, exc_type, exc_value, traceback):
self.stop()
self.join()
if exc_type:
raise exc_type, exc_value, traceback
class Running(object):
def __init__(self, scheduler, name, user='', master=os.getenv('MESOS_MASTER'),
implicit_acknowledge=1, *args, **kwargs):
framework = FrameworkInfo(name=name, user=user, *args, **kwargs)
scheduler = SchedulerProxy(scheduler)
self.driver = MesosSchedulerDriver(scheduler, encode(framework),
master, implicit_acknowledge)
def shutdown(signal, frame):
self.stop()
signal.signal(signal.SIGINT, shutdown)
signal.signal(signal.SIGTERM, shutdown)
atexit.register(self.stop)
def run(self):
return self.driver.run()
def start(self):
status = self.driver.start()
assert status == mesos_pb2.DRIVER_RUNNING
return status
def stop(self):
return self.driver.stop()
def join(self):
return self.driver.join()
def __enter__(self):
self.start()
return self
def __exit__(self, exc_type, exc_value, traceback):
self.stop()
self.join()
if exc_type:
raise exc_type, exc_value, traceback
class Command(BaseCommand):
"""Command that launches the Scale scheduler
"""
help = 'Launches the Scale scheduler'
def add_arguments(self, parser):
parser.add_argument('-m',
'--master',
action='store',
default=settings.MESOS_MASTER,
help='The master to connect to')
def handle(self, *args, **options):
"""See :meth:`django.core.management.base.BaseCommand.handle`.
This method starts the scheduler.
"""
# Register a listener to handle clean shutdowns
signal.signal(signal.SIGTERM, self._onsigterm)
# TODO: clean this up
mesos_master = options.get('master')
logger.info('Scale Scheduler %s', settings.VERSION)
try:
scheduler_zk = settings.SCHEDULER_ZK
except:
scheduler_zk = None
if scheduler_zk is not None:
import socket
from scheduler import cluster_utils
my_id = socket.gethostname()
cluster_utils.wait_for_leader(scheduler_zk, my_id,
self.run_scheduler, mesos_master)
else:
# leader election is disabled
self.run_scheduler(mesos_master)
def run_scheduler(self, mesos_master):
logger.info("I am the leader")
self.scheduler = ScaleScheduler()
framework = mesos_pb2.FrameworkInfo()
framework.user = '' # Have Mesos fill in the current user.
framework.name = os.getenv('DCOS_PACKAGE_FRAMEWORK_NAME', 'Scale')
webserver_address = os.getenv('SCALE_WEBSERVER_ADDRESS')
if webserver_address:
framework.webui_url = webserver_address
logger.info('Connecting to Mesos master at %s', mesos_master)
# TODO(vinod): Make checkpointing the default when it is default on the slave.
if MESOS_CHECKPOINT:
logger.info('Enabling checkpoint for the framework')
framework.checkpoint = True
if MESOS_AUTHENTICATE:
logger.info('Enabling authentication for the framework')
if not DEFAULT_PRINCIPLE:
logger.error(
'Expecting authentication principal in the environment')
sys.exit(1)
if not DEFAULT_SECRET:
logger.error(
'Expecting authentication secret in the environment')
sys.exit(1)
credential = mesos_pb2.Credential()
credential.principal = DEFAULT_PRINCIPLE
credential.secret = DEFAULT_SECRET
self.driver = MesosSchedulerDriver(self.scheduler, framework,
mesos_master, credential)
else:
self.driver = MesosSchedulerDriver(self.scheduler, framework,
mesos_master)
try:
status = 0 if self.driver.run() == mesos_pb2.DRIVER_STOPPED else 1
except:
status = 1
logger.exception('Mesos Scheduler Driver returned an exception')
#Perform a shut down and return any non-zero status
shutdown_status = self._shutdown
status = status or shutdown_status
logger.info('Exiting...')
sys.exit(status)
def _onsigterm(self, signum, _frame):
"""See signal callback registration: :py:func:`signal.signal`.
This callback performs a clean shutdown when a TERM signal is received.
"""
logger.info('Scheduler command terminated due to signal: %i', signum)
self._shutdown()
sys.exit(1)
def _shutdown(self):
"""Performs any clean up required by this command.
:returns: The exit status code based on whether the shutdown operation was clean with no exceptions.
:rtype: int
"""
status = 0
try:
if self.scheduler:
self.scheduler.shutdown()
except:
logger.exception('Failed to properly shutdown Scale scheduler.')
status = 1
try:
if self.driver:
self.driver.stop()
except:
logger.exception('Failed to properly stop Mesos driver.')
status = 1
return status
task = mesos_pb2.TaskInfo()
id = uuid.uuid4()
task.task_id.value = str(id)
task.slave_id.value = offer.slave_id.value
task.name = "task {}".format(str(id))
cpus = task.resources.add()
cpus.name = "cpus"
cpus.type = mesos_pb2.Value.SCALAR
cpus.scalar.value = 0.1
mem = task.resources.add()
mem.name = "mem"
mem.type = mesos_pb2.Value.SCALAR
mem.scalar.value = 32
return task
if __name__ == '__main__':
log("XXX framework started")
framework = mesos_pb2.FrameworkInfo()
framework.user = "******"
framework.name = "nixos-test-framework"
driver = MesosSchedulerDriver(
NixosTestScheduler(),
framework,
sys.argv[1] + ":5050"
)
driver.run()
class Command(BaseCommand):
"""Command that launches the Scale scheduler
"""
option_list = BaseCommand.option_list + (
make_option('-m', '--master', action='store', type='str', default=settings.MESOS_MASTER,
help=('The master to connect to')),
)
help = 'Launches the Scale scheduler'
def handle(self, **options):
"""See :meth:`django.core.management.base.BaseCommand.handle`.
This method starts the scheduler.
"""
# Register a listener to handle clean shutdowns
signal.signal(signal.SIGTERM, self._onsigterm)
# TODO: clean this up
mesos_master = options.get('master')
logger.info('Scale Scheduler %s', settings.VERSION)
try:
scheduler_zk = settings.SCHEDULER_ZK
except:
scheduler_zk = None
if scheduler_zk is not None:
import socket
from scheduler import cluster_utils
my_id = socket.gethostname()
cluster_utils.wait_for_leader(scheduler_zk, my_id, self.run_scheduler, mesos_master)
else:
# leader election is disabled
self.run_scheduler(mesos_master)
def run_scheduler(self, mesos_master):
logger.info("I am the leader")
self.scheduler = ScaleScheduler()
framework = mesos_pb2.FrameworkInfo()
framework.user = '' # Have Mesos fill in the current user.
framework.name = 'Scale'
logger.info('Connecting to Mesos master at %s', mesos_master)
# TODO(vinod): Make checkpointing the default when it is default on the slave.
if MESOS_CHECKPOINT:
logger.info('Enabling checkpoint for the framework')
framework.checkpoint = True
if MESOS_AUTHENTICATE:
logger.info('Enabling authentication for the framework')
if not DEFAULT_PRINCIPLE:
logger.error('Expecting authentication principal in the environment')
sys.exit(1)
if not DEFAULT_SECRET:
logger.error('Expecting authentication secret in the environment')
sys.exit(1)
credential = mesos_pb2.Credential()
credential.principal = DEFAULT_PRINCIPLE
credential.secret = DEFAULT_SECRET
self.driver = MesosSchedulerDriver(self.scheduler, framework, mesos_master, credential)
else:
self.driver = MesosSchedulerDriver(self.scheduler, framework, mesos_master)
try:
status = 0 if self.driver.run() == mesos_pb2.DRIVER_STOPPED else 1
except:
status = 1
logger.exception('Mesos Scheduler Driver returned an exception')
#Perform a shut down and return any non-zero status
shutdown_status = self._shutdown
status = status or shutdown_status
logger.info('Exiting...')
sys.exit(status)
def _onsigterm(self, signum, _frame):
"""See signal callback registration: :py:func:`signal.signal`.
This callback performs a clean shutdown when a TERM signal is received.
"""
logger.info('Scheduler command terminated due to signal: %i', signum)
self._shutdown()
sys.exit(1)
def _shutdown(self):
"""Performs any clean up required by this command.
:returns: The exit status code based on whether the shutdown operation was clean with no exceptions.
:rtype: int
"""
status = 0
try:
if self.scheduler:
self.scheduler.shutdown()
except:
logger.exception('Failed to properly shutdown Scale scheduler.')
status = 1
try:
if self.driver:
self.driver.stop()
except:
logger.exception('Failed to properly stop Mesos driver.')
status = 1
return status
except Exception as e:
raise e
finally:
mesos_lock.release()
db.rhino_tasks.update({'name': task['name']}, {'$set': {'state': 'KILLED'}})
kill_those_that_depend_on(task['name'])
kill_those_that_depend_on(doc['name'])
db.rhino_tasks.update({'mesos_id': status.task_id.value},
{'$set': {'state': state, 'retCode': ret_code, 'message': message}})
if __name__ == '__main__':
try:
web_server_thread = threading.Thread(target=web_server, args=())
web_server_thread.start()
framework = mesos_pb2.FrameworkInfo()
framework.user = "" # Have Mesos fill in the current user.
framework.name = "rhino"
mesos_driver = MesosSchedulerDriver(
AppsomaRhinoScheduler(),
framework,
"zk://" + config['zookeeper_hosts'] + "/mesos"
)
mesos_driver.run()
except KeyboardInterrupt:
print "KeyboardInterrupt"
os.kill(os.getpid(), 9)
class Command(BaseCommand):
'''Command that launches the Scale scheduler
'''
option_list = BaseCommand.option_list + (
make_option('-m', '--master', action='store', type='str', default=settings.MESOS_MASTER,
help=('The master to connect to')),
)
help = 'Launches the Scale scheduler'
def handle(self, **options):
'''See :meth:`django.core.management.base.BaseCommand.handle`.
This method starts the scheduler.
'''
# Register a listener to handle clean shutdowns
signal.signal(signal.SIGTERM, self._onsigterm)
# TODO: clean this up
mesos_master = options.get('master')
logger.info(u'Command starting: scale_scheduler')
logger.info(u' - Master: %s', mesos_master)
executor = mesos_pb2.ExecutorInfo()
executor.executor_id.value = 'scale'
executor.command.value = '%s %s scale_executor' % (settings.PYTHON_EXECUTABLE, settings.MANAGE_FILE)
executor.name = 'Scale Executor (Python)'
self.scheduler = ScaleScheduler(executor)
framework = mesos_pb2.FrameworkInfo()
framework.user = '' # Have Mesos fill in the current user.
framework.name = 'Scale Framework (Python)'
# TODO(vinod): Make checkpointing the default when it is default on the slave.
if MESOS_CHECKPOINT:
logger.info('Enabling checkpoint for the framework')
framework.checkpoint = True
if MESOS_AUTHENTICATE:
logger.info('Enabling authentication for the framework')
if not DEFAULT_PRINCIPLE:
logger.error('Expecting authentication principal in the environment')
sys.exit(1)
if not DEFAULT_SECRET:
logger.error('Expecting authentication secret in the environment')
sys.exit(1)
credential = mesos_pb2.Credential()
credential.principal = DEFAULT_PRINCIPLE
credential.secret = DEFAULT_SECRET
self.driver = MesosSchedulerDriver(self.scheduler, framework, mesos_master, credential)
else:
self.driver = MesosSchedulerDriver(self.scheduler, framework, mesos_master)
status = 0 if self.driver.run() == mesos_pb2.DRIVER_STOPPED else 1
# Perform any required clean up operations like stopping background threads
status = status or self._shutdown()
logger.info(u'Command completed: scale_scheduler')
sys.exit(status)
def _onsigterm(self, signum, _frame):
'''See signal callback registration: :py:func:`signal.signal`.
This callback performs a clean shutdown when a TERM signal is received.
'''
logger.info(u'Scheduler command terminated due to signal: %i', signum)
self._shutdown()
sys.exit(1)
def _shutdown(self):
'''Performs any clean up required by this command.
:returns: The exit status code based on whether the shutdown operation was clean with no exceptions.
:rtype: int
'''
status = 0
try:
if self.scheduler:
self.scheduler.shutdown()
except:
logger.exception('Failed to properly shutdown scale scheduler.')
status = 1
try:
if self.driver:
self.driver.stop()
except:
logger.exception('Failed to properly stop Mesos driver.')
status = 1
return status
elif update.task_id.value == "1":
if update.message == "Command terminated with signal Killed: 9":
print "#### Passed forced shutdown"
else:
print "!!!! Failed forced shutdown"
self.tasksFinished += 1
if self.tasksFinished == TOTAL_TASKS:
driver.stop()
if __name__ == "__main__":
if len(sys.argv) != 2:
print "Usage: %s master" % sys.argv[0]
sys.exit(1)
framework = mesos_pb2.FrameworkInfo()
framework.user = ""
framework.name = "Escalation Framework (Python)"
driver = MesosSchedulerDriver(
EscalationScheduler(),
framework,
sys.argv[1])
status = 0 if driver.run() == mesos_pb2.DRIVER_STOPPED else 1
# Ensure that the driver process terminates.
driver.stop();
sys.exit(status)
mem.type = mesos_pb2.Value.SCALAR
mem.scalar.value = 1
time.sleep(2)
logging.info("Launching task {task} "
"using offer {offer}.".format(task=task.task_id.value,
offer=offer.id.value))
tasks = [task]
driver.launchTasks(offer.id, tasks)
else:
driver.stop()
if __name__ == '__main__':
# make us a framework
framework = mesos_pb2.FrameworkInfo()
framework.user = "******" # Have Mesos fill in the current user.
framework.name = "hello-world"
framework.checkpoint = True
driver = MesosSchedulerDriver(
HelloWorldScheduler(),
framework,
"127.0.0.1:5050/" # assumes running on the master
)
sys.exit(0 if driver.run() == mesos_pb2.DRIVER_STOPPED else 1)