def read_meta():
client = HTTPClient()
for field in AmazonEC2.FIELDS:
try:
value = client.get('%s/%s' % (AmazonEC2.META_URL, field), timeout=0.1, json=False, log=False)
if value is not None:
AmazonEC2.metadata[field] = value
except Exception:
pass
return AmazonEC2.metadata
def __init__(self, foreground=False):
# daemon specific
self.stdin_path = '/dev/null'
if foreground:
self.stdout_path = '/dev/stdout'
self.stderr_path = '/dev/stderr'
else:
self.stdout_path = '/dev/null'
self.stderr_path = '/dev/null'
self.pidfile_path = context.app_config['daemon']['pid']
self.pidfile_timeout = 1
# init
self.client = HTTPClient()
self.containers = {}
self.bridge = None
self.start_time = int(time.time())
self.last_cloud_talk_time = 0
self.is_running = True
class Supervisor(Singleton):
"""
Agent supervisor
Starts dedicated threads for each data source
"""
CONTAINER_CLASS = '%sContainer'
CONTAINER_MODULE = 'amplify.agent.containers.%s.%s'
def __init__(self, foreground=False):
# daemon specific
self.stdin_path = '/dev/null'
if foreground:
self.stdout_path = '/dev/stdout'
self.stderr_path = '/dev/stderr'
else:
self.stdout_path = '/dev/null'
self.stderr_path = '/dev/null'
self.pidfile_path = context.app_config['daemon']['pid']
self.pidfile_timeout = 1
# init
self.client = HTTPClient()
self.containers = {}
self.bridge = None
self.start_time = int(time.time())
self.last_cloud_talk_time = 0
self.is_running = True
def init_containers(self):
"""
Tries to load and create all objects containers specified in config
"""
containers = {}
containers_from_local_config = context.app_config['containers']
for container_name in containers_from_local_config.keys():
try:
container_classname = self.CONTAINER_CLASS % container_name.title()
container_class = loader.import_class(self.CONTAINER_MODULE % (container_name, container_classname))
containers[container_name] = container_class()
context.log.debug('loaded container "%s" from %s' % (container_name, container_class))
except:
context.log.error('failed to load container %s' % container_name, exc_info=True)
return containers
def run(self):
current_thread().name = 'supervisor'
# get initial config from cloud
self.talk_with_cloud()
# run containers
self.containers = self.init_containers()
if not self.containers:
context.log.error('no containers configured, stopping')
return
# run bridge thread
self.bridge = spawn(Bridge().run)
# main cycle
while self.is_running:
context.inc_action_id()
for container in self.containers.itervalues():
container._discover_objects()
container.run_objects()
container.schedule_cloud_commands()
try:
self.talk_with_cloud(top_object=context.top_object.definition)
except AmplifyCriticalException:
pass
self.check_bridge()
time.sleep(5.0)
def stop(self):
self.is_running = False
for container in self.containers.itervalues():
container.stop_objects()
Bridge().flush_metrics()
Bridge().flush_events()
def talk_with_cloud(self, top_object=None):
# TODO: receive commands from cloud
now = int(time.time())
if now <= self.last_cloud_talk_time + 60:
return
# talk to cloud
try:
cloud_response = self.client.post('agent/', data=top_object)
except:
context.log.error('could not connect to cloud', exc_info=True)
raise AmplifyCriticalException()
# update special object configs
changed_containers = []
for obj_config in cloud_response['objects']:
obj = obj_config['object']
obj_type = obj['type']
container = self.containers[obj_type]
obj_id = definition_id(obj)
if container.object_configs.get(obj_id, {}) != obj_config['config']:
container.object_configs[obj_id] = obj_config['config']
changed_containers.append(obj_type)
for obj_type in changed_containers:
self.containers[obj_type].stop_objects()
# TODO work with messages
messages = cloud_response['messages']
# global config changes
config_changed = context.app_config.apply(cloud_response['config'])
if config_changed and self.containers:
context.log.info('config has changed. now running with: %s' % pprint.pformat(context.app_config.config))
for container in self.containers.itervalues():
container.stop_objects()
self.init_containers()
self.last_cloud_talk_time = int(time.time())
def check_bridge(self):
"""
Check containers threads, restart if some failed
"""
bridge_state = 'OK' if not self.bridge.ready() else 'STOP'
# context.log.debug('bridge status: %s' % bridge_state)
if self.bridge.ready and self.bridge.exception:
context.log.debug('bridge exception: %s' % self.bridge.exception)
self.bridge = gevent.spawn(Bridge().run)
def __init__(self):
self.client = HTTPClient()
self.queue = defaultdict(list)
self.first_run = True
class Bridge(Singleton):
def __init__(self):
self.client = HTTPClient()
self.queue = defaultdict(list)
self.first_run = True
def look_around(self):
"""
Checks everything around and make appropriate tree structure
:return: dict of structure
"""
# TODO check docker or OS around
tree = {'system': ['nginx']}
return tree
def run(self):
# TODO: stop after 3 fatal errors
current_thread().name = 'bridge'
context.setup_thread_id()
try:
while True:
m_size_b, m_rss_b = memusage.report()
self.wait()
context.inc_action_id()
self.flush_meta()
if not self.first_run:
self.flush_metrics()
self.flush_events()
self.flush_configs()
else:
self.first_run = False
gc.collect()
m_size_a, m_rss_a = memusage.report()
context.log.debug('mem before: (%s %s), after (%s, %s)' % (m_size_b, m_rss_b, m_size_a, m_rss_a))
if m_rss_a >= m_rss_b:
context.log.debug('RSS MEMORY same or increased, diff %s' % (m_rss_a-m_rss_b))
elif m_size_a >= m_size_b:
context.log.debug('VSIZE MEMORY same or increased, diff %s' % (m_size_a-m_size_b))
except:
context.default_log.error('failed', exc_info=True)
raise
def flush_meta(self):
self._flush(container=context.metad, location='meta/')
def flush_metrics(self):
self._flush(container=context.statsd, location='metrics/')
def flush_events(self):
self._flush(container=context.eventd, location='events/')
def flush_configs(self):
self._flush(container=context.configd, location='nginx/config/')
def _flush(self, container=None, location=None):
# get structure
objects_structure = self.look_around()
# get root object
# there is only one root object!
root_type = objects_structure.keys().pop()
root = container.flush(type=root_type).values()
if not len(root):
root = {'object': context.top_object.definition}
root_only_for_structure = True
else:
root = root[0]
root_only_for_structure = False
# go through children (one level)
root['children'] = []
child_data_found = False
for child_type in objects_structure[root_type]:
for child_data in container.flush(type=child_type).itervalues():
if child_data:
child_data_found = True
root['children'].append(child_data)
self.queue[location].append(root)
if child_data_found or not root_only_for_structure:
context.default_log.debug(root)
try:
self.client.post(location, data=self.queue[location])
self.queue[location] = []
except Exception, e:
exception_name = e.__class__.__name__
context.log.error('failed to push data due to %s' % exception_name)
context.log.debug('additional info:', exc_info=True)
