def subscribe(actor_id, worker_ch):
"""
Main loop for the Actor executor worker. Subscribes to the actor's inbox and executes actor
containers when message arrive. Also subscribes to the worker channel for future communications.
:return:
"""
actor_ch = ActorMsgChannel(actor_id)
t = threading.Thread(target=process_worker_ch, args=(worker_ch, actor_id, actor_ch))
t.start()
print("Worker subscribing to actor channel...")
while keep_running:
update_worker_status(actor_id, worker_ch.name, READY)
try:
msg = actor_ch.get(timeout=2)
except channelpy.ChannelTimeoutException:
continue
print("Received message {}. Starting actor container...".format(str(msg)))
message = msg.pop("msg", "")
try:
stats, logs = execute_actor(actor_id, worker_ch, image, message, msg)
except DockerStartContainerError as e:
print("Got DockerStartContainerError: {}".format(str(e)))
Actor.set_status(actor_id, ERROR)
continue
# add the execution to the actor store
print("Actor container finished successfully. Got stats object:{}".format(str(stats)))
exc_id = Execution.add_execution(actor_id, stats)
Execution.set_logs(exc_id, logs)
def subscribe(tenant, actor_id, worker_id, api_server, client_id,
client_secret, access_token, refresh_token, worker_ch):
"""
Main loop for the Actor executor worker. Subscribes to the actor's inbox and executes actor
containers when message arrive. Also subscribes to the worker channel for future communications.
:return:
"""
logger.debug("Top of subscribe().")
actor_ch = ActorMsgChannel(actor_id)
try:
leave_containers = Config.get('workers', 'leave_containers')
except configparser.NoOptionError:
leave_containers = False
ag = None
if api_server and client_id and client_secret and access_token and refresh_token:
logger.info("Creating agave client.")
verify = get_tenant_verify(tenant)
ag = Agave(api_server=api_server,
token=access_token,
refresh_token=refresh_token,
api_key=client_id,
api_secret=client_secret,
verify=verify)
else:
logger.info("Not creating agave client.")
logger.info("Starting the process worker channel thread.")
t = threading.Thread(target=process_worker_ch,
args=(tenant, worker_ch, actor_id, worker_id,
actor_ch, ag))
t.start()
logger.info("Worker subscribing to actor channel.")
update_worker_status = True
global keep_running
while keep_running:
if update_worker_status:
Worker.update_worker_status(actor_id, worker_id, READY)
update_worker_status = False
try:
msg = actor_ch.get(timeout=2)
except channelpy.ChannelTimeoutException:
continue
except channelpy.ChannelClosedException:
logger.info("Channel closed, worker exiting...")
keep_running = False
sys.exit()
update_worker_status = True
logger.info(
"Received message {}. Starting actor container...".format(msg))
# the msg object is a dictionary with an entry called message and an arbitrary
# set of k:v pairs coming in from the query parameters.
message = msg.pop('message', '')
actor = Actor.from_db(actors_store[actor_id])
execution_id = msg['_abaco_execution_id']
content_type = msg['_abaco_Content_Type']
mounts = actor.mounts
logger.debug("actor mounts: {}".format(mounts))
# for binary data, create a fifo in the configured directory. The configured
# fifo_host_path_dir is equal to the fifo path in the worker container:
fifo_host_path = None
if content_type == 'application/octet-stream':
try:
fifo_host_path_dir = Config.get('workers',
'fifo_host_path_dir')
except (configparser.NoSectionError, configparser.NoOptionError):
logger.error(
"No fifo_host_path configured. Cannot manage binary data.")
Actor.set_status(
actor_id,
ERROR,
msg="Abaco instance not configured for binary data.")
continue
fifo_host_path = os.path.join(fifo_host_path_dir, worker_id,
execution_id)
logger.info("Create fifo at path: {}".format(fifo_host_path))
try:
os.mkfifo(fifo_host_path)
except Exception as e:
logger.error(
"Could not create fifo_path. Exception: {}".format(e))
raise e
# add the fifo as a mount:
mounts.append({
'host_path': fifo_host_path,
'container_path': '/_abaco_binary_data',
'format': 'ro'
})
# the execution object was created by the controller, but we need to add the worker id to it now that we
# know which worker will be working on the execution.
logger.debug("Adding worker_id to execution.")
Execution.add_worker_id(actor_id, execution_id, worker_id)
# privileged dictates whether the actor container runs in privileged mode and if docker daemon is mounted.
privileged = False
if type(actor['privileged']) == bool and actor['privileged']:
privileged = True
logger.debug("privileged: {}".format(privileged))
# retrieve the default environment registered with the actor.
environment = actor['default_environment']
logger.debug("Actor default environment: {}".format(environment))
# construct the user field from the actor's uid and gid:
user = get_container_user(actor)
logger.debug("Final user valiue: {}".format(user))
# overlay the default_environment registered for the actor with the msg
# dictionary
environment.update(msg)
environment['_abaco_access_token'] = ''
environment['_abaco_actor_dbid'] = actor_id
environment['_abaco_actor_id'] = actor.id
environment['_abaco_actor_state'] = actor.state
logger.debug("Overlayed environment: {}".format(environment))
# if we have an agave client, get a fresh set of tokens:
if ag:
try:
ag.token.refresh()
token = ag.token.token_info['access_token']
environment['_abaco_access_token'] = token
logger.info(
"Refreshed the tokens. Passed {} to the environment.".
format(token))
except Exception as e:
logger.error(
"Got an exception trying to get an access token: {}".
format(e))
else:
logger.info(
"Agave client `ag` is None -- not passing access token.")
logger.info("Passing update environment: {}".format(environment))
try:
stats, logs, final_state, exit_code, start_time = execute_actor(
actor_id, worker_id, worker_ch, image, message, user,
environment, privileged, mounts, leave_containers,
fifo_host_path)
except DockerStartContainerError as e:
logger.error("Got DockerStartContainerError: {}".format(e))
Actor.set_status(actor_id, ERROR,
"Error executing container: {}".format(e))
continue
# Add the completed stats to the execution
logger.info(
"Actor container finished successfully. Got stats object:{}".
format(str(stats)))
Execution.finalize_execution(actor_id, execution_id, COMPLETE, stats,
final_state, exit_code, start_time)
logger.info("Added execution: {}".format(execution_id))
# Add the logs to the execution
Execution.set_logs(execution_id, logs)
logger.info("Added execution logs.")
# Update the worker's last updated and last execution fields:
Worker.update_worker_execution_time(actor_id, worker_id)
logger.info("worker time stamps updated.")
def subscribe(tenant,
actor_id,
api_server,
client_id,
client_secret,
access_token,
refresh_token,
worker_ch):
"""
Main loop for the Actor executor worker. Subscribes to the actor's inbox and executes actor
containers when message arrive. Also subscribes to the worker channel for future communications.
:return:
"""
actor_ch = ActorMsgChannel(actor_id)
ag = None
if api_server and client_id and client_secret and access_token and refresh_token:
ag = Agave(api_server=api_server,
token=access_token,
refresh_token=refresh_token,
api_key=client_id,
api_secret=client_secret)
else:
print("Not creating agave client.")
t = threading.Thread(target=process_worker_ch, args=(tenant, worker_ch, actor_id, actor_ch, ag))
t.start()
print("Worker subscribing to actor channel...")
global keep_running
while keep_running:
Worker.update_worker_status(actor_id, worker_ch.name, READY)
try:
msg = actor_ch.get(timeout=2)
except channelpy.ChannelTimeoutException:
continue
except channelpy.ChannelClosedException:
print("Channel closed, worker exiting...")
keep_running = False
sys.exit()
print("Received message {}. Starting actor container...".format(str(msg)))
# the msg object is a dictionary with an entry called message and an arbitrary
# set of k:v pairs coming in from the query parameters.
message = msg.pop('message', '')
actor = Actor.from_db(actors_store[actor_id])
execution_id = msg['_abaco_execution_id']
privileged = False
if actor['privileged'] == 'TRUE':
privileged = True
environment = actor['default_environment']
print("Actor default environment: {}".format(environment))
print("Actor privileged: {}".format(privileged))
# overlay the default_environment registered for the actor with the msg
# dictionary
environment.update(msg)
environment['_abaco_access_token'] = ''
# if we have an agave client, get a fresh set of tokens:
if ag:
try:
ag.token.refresh()
token = ag.token.token_info['access_token']
environment['_abaco_access_token'] = token
print("Refreshed the tokens. Passed {} to the environment.".format(token))
except Exception as e:
print("Got an exception trying to get an access token: {}".format(e))
else:
print("Agave client `ag` is None -- not passing access token.")
print("Passing update environment: {}".format(environment))
try:
stats, logs = execute_actor(actor_id, worker_ch, image, message,
environment, privileged)
except DockerStartContainerError as e:
print("Got DockerStartContainerError: {}".format(str(e)))
Actor.set_status(actor_id, ERROR)
continue
# add the execution to the actor store
print("Actor container finished successfully. Got stats object:{}".format(str(stats)))
Execution.finalize_execution(actor_id, execution_id, COMPLETE, stats)
print("Added execution: {}".format(execution_id))
Execution.set_logs(execution_id, logs)
Worker.update_worker_execution_time(actor_id, worker_ch.name)
def subscribe(tenant, actor_id, worker_id, api_server, client_id,
client_secret, access_token, refresh_token, worker_ch):
"""
Main loop for the Actor executor worker. Subscribes to the actor's inbox and executes actor
containers when message arrive. Also subscribes to the worker channel for future communications.
:return:
"""
logger.debug("Top of subscribe().")
actor_ch = ActorMsgChannel(actor_id)
ag = None
if api_server and client_id and client_secret and access_token and refresh_token:
logger.info("Creating agave client.")
ag = Agave(api_server=api_server,
token=access_token,
refresh_token=refresh_token,
api_key=client_id,
api_secret=client_secret)
else:
logger.info("Not creating agave client.")
logger.info("Starting the process worker channel thread.")
t = threading.Thread(target=process_worker_ch,
args=(tenant, worker_ch, actor_id, worker_id,
actor_ch, ag))
t.start()
logger.info("Worker subscribing to actor channel.")
update_worker_status = True
global keep_running
while keep_running:
if update_worker_status:
Worker.update_worker_status(actor_id, worker_id, READY)
update_worker_status = False
try:
msg = actor_ch.get(timeout=2)
except channelpy.ChannelTimeoutException:
continue
except channelpy.ChannelClosedException:
logger.info("Channel closed, worker exiting...")
keep_running = False
sys.exit()
update_worker_status = True
logger.info(
"Received message {}. Starting actor container...".format(msg))
# the msg object is a dictionary with an entry called message and an arbitrary
# set of k:v pairs coming in from the query parameters.
message = msg.pop('message', '')
actor = Actor.from_db(actors_store[actor_id])
execution_id = msg['_abaco_execution_id']
# the execution object was created by the controller, but we need to add the worker id to it now that we
# know which worker will be working on the execution.
logger.debug("Adding worker_id to execution.")
Execution.add_worker_id(actor_id, execution_id, worker_id)
# privileged dictates whether the actor container runs in privileged mode and if docker daemon is mounted.
privileged = False
if actor['privileged'] == 'TRUE':
privileged = True
logger.debug("privileged: {}".format(privileged))
# retrieve the default environment registered with the actor.
environment = actor['default_environment']
logger.debug("Actor default environment: {}".format(environment))
# overlay the default_environment registered for the actor with the msg
# dictionary
environment.update(msg)
environment['_abaco_access_token'] = ''
environment['_abaco_actor_dbid'] = actor_id
environment['_abaco_actor_id'] = actor.id
environment['_abaco_actor_state'] = actor.state
logger.debug("Overlayed environment: {}".format(environment))
# if we have an agave client, get a fresh set of tokens:
if ag:
try:
ag.token.refresh()
token = ag.token.token_info['access_token']
environment['_abaco_access_token'] = token
logger.info(
"Refreshed the tokens. Passed {} to the environment.".
format(token))
except Exception as e:
logger.error(
"Got an exception trying to get an access token: {}".
format(e))
else:
logger.info(
"Agave client `ag` is None -- not passing access token.")
logger.info("Passing update environment: {}".format(environment))
try:
stats, logs, final_state, exit_code = execute_actor(
actor_id, worker_id, worker_ch, image, message, environment,
privileged)
except DockerStartContainerError as e:
logger.error("Got DockerStartContainerError: {}".format(str(e)))
Actor.set_status(actor_id, ERROR)
continue
# Add the completed stats to the execution
logger.info(
"Actor container finished successfully. Got stats object:{}".
format(str(stats)))
Execution.finalize_execution(actor_id, execution_id, COMPLETE, stats,
final_state, exit_code)
logger.info("Added execution: {}".format(execution_id))
# Add the logs to the execution
Execution.set_logs(execution_id, logs)
logger.info("Added execution logs.")
# Update the worker's last updated and last execution fields:
Worker.update_worker_execution_time(actor_id, worker_id)
logger.info("worker time stamps updated.")
def subscribe(tenant, actor_id, worker_id, api_server, client_id,
client_secret, access_token, refresh_token, worker_ch):
"""
Main loop for the Actor executor worker. Subscribes to the actor's inbox and executes actor
containers when message arrive. Also subscribes to the worker channel for future communications.
:return:
"""
actor_ch = ActorMsgChannel(actor_id)
ag = None
if api_server and client_id and client_secret and access_token and refresh_token:
ag = Agave(api_server=api_server,
token=access_token,
refresh_token=refresh_token,
api_key=client_id,
api_secret=client_secret)
else:
print("Not creating agave client.")
t = threading.Thread(target=process_worker_ch,
args=(tenant, worker_ch, actor_id, worker_id,
actor_ch, ag))
t.start()
print("Worker subscribing to actor channel...")
global keep_running
while keep_running:
Worker.update_worker_status(actor_id, worker_id, READY)
try:
msg = actor_ch.get(timeout=2)
except channelpy.ChannelTimeoutException:
continue
except channelpy.ChannelClosedException:
print("Channel closed, worker exiting...")
keep_running = False
sys.exit()
print("Received message {}. Starting actor container...".format(
str(msg)))
# the msg object is a dictionary with an entry called message and an arbitrary
# set of k:v pairs coming in from the query parameters.
message = msg.pop('message', '')
actor = Actor.from_db(actors_store[actor_id])
execution_id = msg['_abaco_execution_id']
Execution.add_worker_id(actor_id, execution_id, worker_id)
privileged = False
if actor['privileged'] == 'TRUE':
privileged = True
environment = actor['default_environment']
print("Actor default environment: {}".format(environment))
print("Actor privileged: {}".format(privileged))
# overlay the default_environment registered for the actor with the msg
# dictionary
environment.update(msg)
environment['_abaco_access_token'] = ''
environment['_abaco_actor_dbid'] = actor_id
environment['_abaco_actor_id'] = actor.id
environment['_abaco_actor_state'] = actor.state
# if we have an agave client, get a fresh set of tokens:
if ag:
try:
ag.token.refresh()
token = ag.token.token_info['access_token']
environment['_abaco_access_token'] = token
print("Refreshed the tokens. Passed {} to the environment.".
format(token))
except Exception as e:
print("Got an exception trying to get an access token: {}".
format(e))
else:
print("Agave client `ag` is None -- not passing access token.")
print("Passing update environment: {}".format(environment))
try:
stats, logs, final_state, exit_code = execute_actor(
actor_id, worker_id, worker_ch, image, message, environment,
privileged)
except DockerStartContainerError as e:
print("Got DockerStartContainerError: {}".format(str(e)))
Actor.set_status(actor_id, ERROR)
continue
# add the execution to the actor store
print("Actor container finished successfully. Got stats object:{}".
format(str(stats)))
Execution.finalize_execution(actor_id, execution_id, COMPLETE, stats,
final_state, exit_code)
print("Added execution: {}".format(execution_id))
Execution.set_logs(execution_id, logs)
Worker.update_worker_execution_time(actor_id, worker_id)
