def handle(self):
print "SERVER REQUEST", self.__class__, "ME: ", self.request.getsockname()
print "FROM :", self.request.getpeername()
message = recv_unicode_netstring(self.rfile)
print message
pl = PollingObject()
parsed = self.server.mtool.unpack_msg(message)
if parsed.type == "find_eeg_experiments" or parsed.type == "find_eeg_amplifiers":
pull_addr = "tcp://" + socket.gethostname() + ":" + str(self.server.pull_port)
parsed.client_push_address = pull_addr
srv_sock = self.make_srv_sock()
try:
send_msg(srv_sock, parsed.SerializeToString())
response, det = pl.poll_recv(srv_sock, timeout=5000)
finally:
srv_sock.close()
print "passed msg and got result: ", response
if not response:
self.bad_response(self.wfile, det)
return
if parsed.type == "find_eeg_experiments" or parsed.type == "find_eeg_amplifiers":
response, det = pl.poll_recv(self.server.pull_sock, timeout=20000)
if not response:
self.bad_response(self.wfile, det)
return
data = make_unicode_netstring(response)
self.wfile.write(data)
def handle(self):
print "SERVER REQUEST", self.__class__, "ME: ", self.request.getsockname(
)
print "FROM :", self.request.getpeername()
message = recv_unicode_netstring(self.rfile)
print message
pl = PollingObject()
parsed = self.server.mtool.unpack_msg(message)
if parsed.type == 'find_eeg_experiments' or parsed.type == 'find_eeg_amplifiers':
pull_addr = 'tcp://' + socket.gethostname() + ':' + str(
self.server.pull_port)
parsed.client_push_address = pull_addr
srv_sock = self.make_srv_sock()
try:
send_msg(srv_sock, parsed.SerializeToString())
response, det = pl.poll_recv(srv_sock, timeout=5000)
finally:
srv_sock.close()
print "passed msg and got result: ", response
if not response:
self.bad_response(self.wfile, det)
return
if parsed.type == 'find_eeg_experiments' or parsed.type == 'find_eeg_amplifiers':
response, det = pl.poll_recv(self.server.pull_sock, timeout=20000)
if not response:
self.bad_response(self.wfile, det)
return
data = make_unicode_netstring(response)
self.wfile.write(data)
def _publisher_thread(self, pub_addrs, pull_address, push_addr):
#FIXME aaaaahhh pub_addresses are set here, not in the main thread
# (which reads them in _register method)
pub_sock, self.pub_addresses = self._init_socket(
pub_addrs, zmq.PUB)
pull_sock = self.ctx.socket(zmq.PULL)
pull_sock.bind(pull_address)
push_sock = self.ctx.socket(zmq.PUSH)
push_sock.connect(push_addr)
send_msg(push_sock, u'1')
po = PollingObject()
while not self._stop_publishing:
try:
to_publish, det = po.poll_recv(pull_sock, 500)
if to_publish:
send_msg(pub_sock, to_publish)
except:
#print self.name, '.Publisher -- STOP.'
break
# self.logger.info( "close sock %s %s", pub_addrs, pub_sock)
pub_sock.close()
pull_sock.close()
push_sock.close()
def _publisher_thread(self, pub_addrs, pull_address, push_addr):
# FIXME aaaaahhh pub_addresses are set here, not in the main thread
# (which reads them in _register method)
pub_sock, self.pub_addresses = self._init_socket(
pub_addrs, zmq.PUB)
pull_sock = self.ctx.socket(zmq.PULL)
pull_sock.bind(pull_address)
push_sock = self.ctx.socket(zmq.PUSH)
push_sock.connect(push_addr)
send_msg(push_sock, b'1')
po = PollingObject()
while not self._stop_publishing:
try:
to_publish, det = po.poll_recv(pull_sock, 500)
if to_publish:
send_msg(pub_sock, to_publish)
except:
# print self.name, '.Publisher -- STOP.'
break
# self.logger.info( "close sock %s %s", pub_addrs, pub_sock)
pub_sock.close()
pull_sock.close()
push_sock.close()
def __init__(self,
proc_description,
reg_timeout_desc=None,
monitoring_optflags=PING,
logger=None):
self.desc = proc_description
self.must_register = reg_timeout_desc is not None
self._status_lock = threading.RLock()
self._status = UNKNOWN if self.must_register else RUNNING
self._status_details = None
self.ping_it = monitoring_optflags & PING
self.check_returncode = monitoring_optflags & RETURNCODE if \
self.desc.pid is not None else False
self.logger = logger or get_logger(
'subprocess_monitor' + '-' + self.desc.name + '-' +
str(self.desc.pid),
stream_level='info')
self.set_registration_timeout_handler(reg_timeout_desc)
self.registration_data = None
self._stop_monitoring = False
self._ping_thread = None
self._ping_retries = 8
self._returncode_thread = None
self._mtool = OBCIMessageTool(message_templates)
self._ctx = None
self.rq_sock = None
self._poller = PollingObject()
self.delete = False
def __init__(self, zmq_ctx, uuid, logger=None, obci_dns=None):
self._processes = {}
self._ctx = zmq_ctx
self.uuid = uuid
self.logger = logger or get_logger('subprocess_monitor',
stream_level='warning')
self.obci_dns = obci_dns
self._mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self._proc_lock = threading.RLock()
def __init__(self, srv_addrs, ctx, client_push_address, nearby_servers):
self.ctx = ctx
self.server_req_socket = self.ctx.socket(zmq.REQ)
for addr in srv_addrs:
self.server_req_socket.connect(addr)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.nearby_servers = nearby_servers
self._amplified_cache = {}
def handle(self):
message = recv_unicode_netstring(self.rfile)
srv_sock = self.make_srv_sock()
try:
send_msg(srv_sock, message)
pl = PollingObject()
response, det = pl.poll_recv(srv_sock, timeout=5000)
finally:
srv_sock.close()
if not response:
self.bad_response(self.wfile, det)
self.rfile.write(make_unicode_netstring(response))
def handle(self):
message = recv_unicode_netstring(self.rfile)
srv_sock = self.make_srv_sock()
try:
send_msg(srv_sock, message)
pl = PollingObject()
response, det = pl.poll_recv(srv_sock, timeout=5000)
finally:
srv_sock.close()
if not response:
self.bad_response(self.wfile, det)
self.rfile.write(make_unicode_netstring(response))
def __init__(self, proc_description,
reg_timeout_desc=None,
monitoring_optflags=PING,
logger=None):
self.desc = proc_description
self.must_register = reg_timeout_desc is not None
self._status_lock = threading.RLock()
self._status = UNKNOWN if self.must_register else RUNNING
self._status_details = None
self.ping_it = monitoring_optflags & PING
self.check_returncode = monitoring_optflags & RETURNCODE if \
self.desc.pid is not None else False
self.logger = logger or get_logger(
'subprocess_monitor'+'-'+self.desc.name+'-'+str(self.desc.pid),
stream_level='info')
self.set_registration_timeout_handler(reg_timeout_desc)
self.registration_data = None
self._stop_monitoring = False
self._ping_thread = None
self._ping_retries = 8
self._returncode_thread = None
self._mtool = OBCIMessageTool(message_templates)
self._ctx = None
self.rq_sock = None
self._poller = PollingObject()
self.delete = False
def __init__(self, srv_addrs, ctx, client_push_address, nearby_servers):
self.ctx = ctx
self.server_req_socket = self.ctx.socket(zmq.REQ)
for addr in srv_addrs:
self.server_req_socket.connect(addr)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.nearby_servers = nearby_servers
self._amplified_cache = {}
def __init__(self, zmq_ctx, uuid, logger=None, obci_dns=None):
self._processes = {}
self._ctx = zmq_ctx
self.uuid = uuid
self.logger = logger or get_logger('subprocess_monitor',
stream_level='warning')
self.obci_dns = obci_dns
self._mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self._proc_lock = threading.RLock()
def kill(self):
# send "kill" to the process or kill request to its supervisor?
self.stop_monitoring()
if not self._ctx:
self._ctx = zmq.Context()
rq_sock = self._ctx.socket(zmq.REQ)
try:
rq_sock.connect(self.rq_address)
mtool = OBCIMessageTool(message_templates)
poller = PollingObject()
send_msg(rq_sock, mtool.fill_msg("kill_process", pid=self.pid, machine=self.machine_ip))
res, _ = poller.poll_recv(rq_sock, timeout=5000)
finally:
rq_sock.close()
if res:
res = mtool.unpack_msg(res)
print "Response to kill request: ", res
with self._status_lock:
self._status = TERMINATED
def stringReceived(self, string):
req_sock = self.factory.ctx.socket(zmq.REQ)
req_sock.connect(self.factory.zmq_rep_addr)
try:
req = unicode(string, encoding='utf-8')
print "twisted got:", req
bad = False
try:
parsed = self.factory.mtool.unpack_msg(req)
except ValueError:
bad = True
if not bad:
if parsed.type in self.factory.long_rqs:
sock, port = self.factory.long_rqs[parsed.type]
pull_addr = 'tcp://' + socket.gethostname() + ':' + str(
port)
parsed.client_push_address = pull_addr
send_msg(req_sock, parsed.SerializeToString())
else:
send_msg(req_sock, req)
pl = PollingObject()
msg, det = pl.poll_recv(req_sock, timeout=5000)
finally:
req_sock.close()
if not msg:
msg = self.factory.mtool.fill_msg("rq_error", details=det)
if not bad:
if parsed.type in self.factory.long_rqs:
sock, port = self.factory.long_rqs[parsed.type]
msg, det = pl.poll_recv(sock, timeout=20000)
if not msg:
msg = self.factory.mtool.fill_msg("rq_error", details=det)
return
encmsg = msg.encode('utf-8')
self.sendString(encmsg)
reactor.callFromThread(self.sendString, encmsg)
def stringReceived(self, string):
req_sock = self.factory.ctx.socket(zmq.REQ)
req_sock.connect(self.factory.zmq_rep_addr)
try:
req = unicode(string, encoding='utf-8')
print "twisted got:", req
bad = False
try:
parsed = self.factory.mtool.unpack_msg(req)
except ValueError:
bad = True
if not bad:
if parsed.type in self.factory.long_rqs:
sock, port = self.factory.long_rqs[parsed.type]
pull_addr = 'tcp://' + socket.gethostname() + ':' + str(port)
parsed.client_push_address = pull_addr
send_msg(req_sock, parsed.SerializeToString())
else:
send_msg(req_sock, req)
pl = PollingObject()
msg, det = pl.poll_recv(req_sock, timeout=5000)
finally:
req_sock.close()
if not msg:
msg = self.factory.mtool.fill_msg("rq_error", details=det)
if not bad:
if parsed.type in self.factory.long_rqs:
sock, port = self.factory.long_rqs[parsed.type]
msg, det = pl.poll_recv(sock, timeout=20000)
if not msg:
msg = self.factory.mtool.fill_msg("rq_error", details=det)
return
encmsg = msg.encode('utf-8')
self.sendString(encmsg)
reactor.callFromThread(self.sendString, encmsg)
def __init__(self, server_addresses, zmq_context=None):
self.ctx = zmq_context if zmq_context else zmq.Context()
self.server_addresses = server_addresses
self.server_req_socket = None
self.init_server_socket(server_addresses)
# self = zmq.Poller()
# self.register(self.server_req_socket, zmq.POLLIN)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.dns = net.DNS()
def __init__(self, server_addresses, zmq_context=None):
self.ctx = zmq_context if zmq_context else zmq.Context()
self.server_addresses = server_addresses
self.server_req_socket = None
self.init_server_socket(server_addresses)
#self = zmq.Poller()
#self.register(self.server_req_socket, zmq.POLLIN)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.dns = net.DNS()
def __init__(self,
server_address,
handler_class,
bind_and_activate=True,
zmq_ctx=None,
zmq_rep_addr=None):
daemon_threads = True
server_timeout = 45
SocketServer.TCPServer.__init__(self, server_address, handler_class,
bind_and_activate)
self.mtool = OBCIMessageTool(message_templates)
self.pl = PollingObject()
self.ctx = zmq_ctx
self.rep_addr = zmq_rep_addr
def find_eeg_experiments_and_push_results(ctx, srv_addrs, rq_message,
nearby_servers):
LOGGER = logger.get_logger("eeg_experiment_finder", "info")
finder = EEGExperimentFinder(srv_addrs, ctx,
rq_message.client_push_address,
nearby_servers)
exps = finder.find_amplified_experiments()
mpoller = PollingObject()
checked = rq_message.checked_srvs
if not isinstance(checked, list):
checked = []
nrb = {}
for uid, srv in nearby_servers.snapshot().iteritems():
if srv.ip not in checked:
nrb[uid] = srv
if not checked and nearby_servers.dict_snapshot():
my_addr = nearby_servers.ip(hostname=socket.gethostname())
LOGGER.info("checking other servers")
print[(srv.hostname, srv.ip) for srv in nrb.values()]
ip_list = [srv.ip for srv in nrb.values() if \
srv.ip != my_addr]
LOGGER.info("number of servers to query: " + str(len(ip_list)))
exps += _gather_other_server_results(ctx, my_addr, ip_list)
else:
LOGGER.info("not checking other servers")
LOGGER.info("return to: " + rq_message.client_push_address)
to_client = ctx.socket(zmq.PUSH)
to_client.connect(rq_message.client_push_address)
send_msg(
to_client,
finder.mtool.fill_msg('eeg_experiments',
sender_ip=socket.gethostname(),
experiment_list=exps))
LOGGER.info("sent exp data... " + str(exps)[:500] + ' [...]')
time.sleep(0.1)
def __init__(self, preset_data=None, launcher_data=None, ctx=None):
self.preset_data = preset_data
self.launch_file = None
self.launcher_data = launcher_data
self.name = None
self.info = ""
self.public_params = []
self.origin_machine = ''
self.unsupervised_peers = {}
self.old_uid = None
self.ctx = ctx
self.exp_req = None
self.mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self.category = DEFAULT_CATEGORY
if preset_data is not None:
self.setup_from_preset(preset_data)
elif launcher_data is not None:
self.setup_from_launcher(launcher_data)
super(ExperimentEngineInfo, self).__init__()
def __init__(self, preset_data=None, launcher_data=None, ctx=None):
self.preset_data = preset_data
self.launch_file = None
self.launcher_data = launcher_data
self.name = None
self.info = ""
self.public_params = []
self.origin_machine = ''
self.unsupervised_peers = {}
self.old_uid = None
self.ctx = ctx
self.exp_req = None
self.mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self.category = DEFAULT_CATEGORY
if preset_data is not None:
self.setup_from_preset(preset_data)
elif launcher_data is not None:
self.setup_from_launcher(launcher_data)
super(ExperimentEngineInfo, self).__init__()
class Process(object):
def __init__(self,
proc_description,
reg_timeout_desc=None,
monitoring_optflags=PING,
logger=None):
self.desc = proc_description
self.must_register = reg_timeout_desc is not None
self._status_lock = threading.RLock()
self._status = UNKNOWN if self.must_register else RUNNING
self._status_details = None
self.ping_it = monitoring_optflags & PING
self.check_returncode = monitoring_optflags & RETURNCODE if \
self.desc.pid is not None else False
self.logger = logger or get_logger(
'subprocess_monitor' + '-' + self.desc.name + '-' +
str(self.desc.pid),
stream_level='info')
self.set_registration_timeout_handler(reg_timeout_desc)
self.registration_data = None
self._stop_monitoring = False
self._ping_thread = None
self._ping_retries = 8
self._returncode_thread = None
self._mtool = OBCIMessageTool(message_templates)
self._ctx = None
self.rq_sock = None
self._poller = PollingObject()
self.delete = False
@property
def machine_ip(self):
return self.desc.machine_ip
@property
def pid(self):
return self.desc.pid
@property
def path(self):
return self.desc.path
@property
def proc_type(self):
return self.desc.proc_type
@property
def name(self):
return self.desc.name
def status(self):
with self._status_lock:
return self._status, self._status_details
def set_registration_timeout_handler(self, reg_timeout_desc):
with self._status_lock:
self._status = UNKNOWN
self._status_details = None
self.must_register = reg_timeout_desc is not None
self.reg_timeout_desc = reg_timeout_desc
self.reg_timer = None if not self.must_register else \
self.new_timer(self.reg_timeout_desc, REG_TIMER)
if self.must_register:
self.reg_timer.start()
def is_local(self):
raise NotImplementedError()
def timeout_handler(self, custom_method, args, type_):
self._do_handle_timeout(type_)
custom_method(*args)
def _do_handle_timeout(self, type_):
raise NotImplementedError()
def new_timer(self, tim_desc, type_):
return threading.Timer(
tim_desc.timeout, self.timeout_handler,
[tim_desc.timeout_method, tim_desc.timeout_args, type_])
def registered(self, reg_data):
if self.reg_timer is not None:
self.reg_timer.cancel()
self.logger.info("{0} [{1}] REGISTERED!!! {2}".format(
self.name, self.proc_type, reg_data.machine_ip))
# print "ping:", self.ping_it, "ret:", self.check_returncode
with self._status_lock:
self._status = RUNNING
# TODO validate registration data
self.registration_data = reg_data
self.logger.info("reg_data" + str(vars(reg_data)))
if self.ping_it:
if not self._ctx:
self._ctx = zmq.Context()
self.rq_sock = self._ctx.socket(zmq.REQ)
for addr in reg_data.rep_addrs:
if reg_data.machine_ip != socket.gethostname() and\
net.addr_is_local(addr):
continue
self.logger.debug(self.name + "connecting to " + addr)
self.rq_sock.connect(addr)
def stop_monitoring(self):
if self.reg_timer:
self.reg_timer.cancel()
self.reg_timer = None
self._stop_monitoring = True
if self._ping_thread is not None:
self.logger.info("%s, %s, %s", self.proc_type, self.name,
"Joining ping thread")
self._ping_thread.join()
if self._returncode_thread is not None:
self.logger.info("%s %s %s", self.proc_type, self.name,
"joining returncode thread")
self._returncode_thread.join()
self.logger.info("monitor for: %s, %s, %s", self.proc_type, self.name,
" ...monitoring threads stopped.")
def finished(self):
finished = True
if self._ping_thread is not None:
finished = not self._ping_thread.is_alive()
if self._returncode_thread is not None:
finished = finished and not self._returncode_thread.is_alive()
return finished
def process_is_running(self):
running = True
if self._ping_thread is not None:
running = self._ping_thread.is_alive()
if self._returncode_thread is not None:
running = running and self._returncode_thread.is_alive()
return running
def start_monitoring(self):
if self.ping_it:
self._ping_thread = threading.Thread(target=self.ping_monitor,
args=())
self._ping_thread.daemon = True
self._ping_thread.start()
if self.check_returncode:
self._returncode_thread = threading.Thread(
target=self.returncode_monitor, args=())
self._returncode_thread.daemon = True
self._returncode_thread.start()
def ping_monitor(self):
is_alive = True
try:
while not self._stop_monitoring and is_alive:
time.sleep(2)
if self.rq_sock is not None:
send_msg(self.rq_sock, self._mtool.fill_msg('ping'))
result = None
while self._ping_retries and not result and not self._stop_monitoring:
result, det = self._poller.poll_recv(
socket=self.rq_sock, timeout=1500)
if not result and not self._stop_monitoring:
self.logger.info("%s %s %s", self.proc_type, self.name,
"NO RESPONSE TO PING!")
with self._status_lock:
if self._status not in [FAILED, FINISHED]:
self._status = NON_RESPONSIVE
self._status_details = 'ping response timeout'
print "status:", self._status
is_alive = False
finally:
if self.rq_sock is not None:
self.rq_sock.close(linger=0)
def returncode_monitor(self):
raise NotImplementedError()
def kill(self):
raise NotImplementedError()
def mark_delete(self):
with self._status_lock:
self.delete = True
def marked_delete(self):
with self._status_lock:
return self.delete
#!/usr/bin/python3
import json
import zmq
import sys
from obci.control.common.message import OBCIMessageTool, send_msg, PollingObject
from obci.control.launcher.launcher_messages import message_templates
from obci.control.common.obci_control_settings import PORT_RANGE
import obci.control.common.net_tools as net
if __name__ == '__main__':
mtool = OBCIMessageTool(message_templates)
pl = PollingObject()
# ifname = net.server_ifname()
my_addr = 'tcp://' + 'localhost'
ctx = zmq.Context()
server_req = ctx.socket(zmq.REQ)
server_req.connect(my_addr + ':' + net.server_rep_port())
exp_info_pull = ctx.socket(zmq.PULL)
port = exp_info_pull.bind_to_random_port('tcp://*',
min_port=PORT_RANGE[0],
max_port=PORT_RANGE[1], max_tries=500)
class SubprocessMonitor(object):
def __init__(self, zmq_ctx, uuid, logger=None, obci_dns=None):
self._processes = {}
self._ctx = zmq_ctx
self.uuid = uuid
self.logger = logger or get_logger('subprocess_monitor',
stream_level='warning')
self.obci_dns = obci_dns
self._mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self._proc_lock = threading.RLock()
def not_running_processes(self):
status = {}
with self._proc_lock:
for key, proc in self._processes.iteritems():
st = proc.status()
if st[0] in [FINISHED, FAILED, TERMINATED] and not proc.marked_delete():
status[key] = st
return status
def unknown_status_processes(self):
with self._proc_lock:
return [proc for proc in self._processes.values()\
if proc.status()[0] == UNKNOWN]
def process(self, machine_ip, pid):
with self._proc_lock:
return self._processes.get((machine_ip, pid), None)
def killall(self, force=False):
with self._proc_lock:
for proc in self._processes.values():
kill_method = proc.kill if not force else proc.kill_with_force
if proc.status()[0] not in [FINISHED, FAILED, TERMINATED]:
kill_method()
def delete(self, machine, pid):
proc = self._processes.get((machine, pid), None)
if proc is None:
raise Exception("Process not found: " + str((machine, pid)))
if not proc.running():
del self._processes[(machine, pid)]
return True
else:
self.logger.error("Process is running, will not delete! " + str((machine, pid)))
return False
def delete_all(self):
with self._proc_lock:
for proc in self._processes.values():
del proc
self._processes = {}
def stop_monitoring(self):
with self._proc_lock:
for proc in self._processes.values():
proc.stop_monitoring()
def _launch_args(self, path, args):
#TODO fix interpreter calls // only python is supported
if path.endswith('.py'):
launch_args = PYTHON_CALL +[path] + args
else:
launch_args = [path] + args
return launch_args
def _stdio_actions(self, io_flags):
out = subprocess.PIPE if io_flags & STDOUT else None
if io_flags & STDERR:
err = subprocess.PIPE
elif out is not None:
err = subprocess.STDOUT
else: err = None
stdin = subprocess.PIPE if io_flags & STDIN else None
return (out, err, stdin)
def _local_launch(self, launch_args, stdio_actions, env):
ON_POSIX = 'posix' in sys.builtin_module_names
out, err, stdin = stdio_actions
try:
if sys.platform == "win32":
crflags = subprocess.CREATE_NEW_PROCESS_GROUP
popen_obj = subprocess.Popen(launch_args,
stdout=out, stderr=err, stdin=stdin,
bufsize=1, close_fds=ON_POSIX, env=env,
creationflags=crflags)
else:
popen_obj = subprocess.Popen(launch_args,
stdout=out, stderr=err, stdin=stdin,
bufsize=1, close_fds=ON_POSIX, env=env)
details = "Popen constructor finished for " +\
str(launch_args[:3]) + "(...)"
self.logger.info(details)
return popen_obj, details
except OSError as e:
details = "Unable to spawn process {0} [{1}]".format(launch_args, e.args)
self.logger.error(details)
return None, details
except ValueError as e:
details = "Unable to spawn process (bad arguments) {0} [{1}]".format(launch_args, e.args)
self.logger.error(details)
return None, details
except Exception as e:
details = "Process launch Error: " + str(e) + str(e.args) + str(vars(e))
self.logger.error(details)
return None, details
def new_local_process(self, path, args, proc_type='', name='',
capture_io= STDOUT | STDIN,
stdout_log=None,
stderr_log=None,
register_timeout_desc=None,
monitoring_optflags=RETURNCODE | PING,
machine_ip=None,
env=None):
launch_args = self._launch_args(path, args)
self.logger.debug(proc_type + " local path: " + path)
machine = machine_ip if machine_ip else socket.gethostname()
std_actions = self._stdio_actions(capture_io)
timeout_desc = register_timeout_desc
self.logger.debug('process launch arg list: %s', launch_args)
popen_obj, details = self._local_launch(launch_args, std_actions, env)
if popen_obj is None:
return None, details
if popen_obj.returncode is not None:
det = "opened process already terminated" + popen_obj.communicate()
self.logger.warning(det)
if not name:
name = os.path.basename(path)
process_desc = ProcessDescription(proc_type=proc_type,
name=name,
path=path,
args=args,
machine_ip=machine,
pid=popen_obj.pid)
# io_handler will be None if no stdio is captured
io_handler = start_stdio_handler(popen_obj, std_actions,
':'.join([machine, path, name]),
stdout_log, stderr_log)
new_proc = LocalProcess(process_desc, popen_obj, io_handler=io_handler,
reg_timeout_desc=timeout_desc,
monitoring_optflags=monitoring_optflags,
logger=self.logger)
if monitoring_optflags & PING:
new_proc._ctx = self._ctx
with self._proc_lock:
self._processes[(machine, popen_obj.pid)] = new_proc
new_proc.start_monitoring()
return new_proc, None
def new_remote_process(self, path, args, proc_type, name,
machine_ip, conn_addr,
capture_io= STDOUT | STDIN,
stdout_log=None,
stderr_log=None,
register_timeout_desc=None,
monitoring_optflags=PING):
"""Send a request to conn_addr for a process launch. By default
the process will be monitored with ping requests and locally by the
remote peer."""
timeout_desc = register_timeout_desc
rq_message = self._mtool.fill_msg('launch_process',
path=path,
args=args, proc_type=proc_type,
name=name,
machine_ip=machine_ip,
capture_io=capture_io,
stdout_log=stdout_log,
stderr_log=stderr_log)
rq_sock = self._ctx.socket(zmq.REQ)
try:
rq_sock.connect(conn_addr)
except zmq.ZMQError as e:
det = "Could not connect to {0}, err: {1}, {2}".format(
conn_addr, e, e.args)
self.logger.error(det)
return None, det
self.logger.info("SENDING LAUNCH REQUEST {0} {1} {2} {3}".format(
machine_ip, _DEFAULT_TIMEOUT_MS, 'ms', conn_addr))
send_msg(rq_sock, rq_message)
result, details = self.poller.poll_recv(rq_sock, _DEFAULT_TIMEOUT_MS)
rq_sock.close()
if not result:
details = details + " [address was: {0}]".format(conn_addr)
self.logger.error(details)
return None, details
else:
result = self._mtool.unpack_msg(result)
if result.type == 'rq_error':
det = "REQUEST FAILED" + str(result.err_code) + ':' + str(result.details)
self.logger.error(det)
return None, det
elif result.type == 'launched_process_info':
self.logger.info("REQUEST SUCCESS %s", result.dict())
process_desc = ProcessDescription(proc_type=result.proc_type,
name=result.name,
path=result.path,
args=args,
machine_ip=result.machine,
pid=result.pid)
new_proc = RemoteProcess(process_desc, conn_addr,
reg_timeout_desc=timeout_desc,
monitoring_optflags=monitoring_optflags,
logger=self.logger)
if monitoring_optflags & PING:
new_proc._ctx = self._ctx
with self._proc_lock:
self._processes[(result.machine, result.pid)] = new_proc
new_proc.start_monitoring()
return new_proc, None
# -*- coding: utf-8 -*-
import json
import zmq
import sys
import socket
from obci.control.common.message import OBCIMessageTool, send_msg, recv_msg, PollingObject
from obci.control.launcher.launcher_messages import message_templates, error_codes
from obci.control.common.obci_control_settings import PORT_RANGE
import obci.control.common.net_tools as net
if __name__ == '__main__':
mtool = OBCIMessageTool(message_templates)
pl = PollingObject()
# ifname = net.server_ifname()
my_addr = 'tcp://' + 'localhost'
ctx = zmq.Context()
server_req = ctx.socket(zmq.REQ)
server_req.connect(my_addr + ':' + net.server_rep_port())
exp_info_pull = ctx.socket(zmq.PULL)
port = exp_info_pull.bind_to_random_port('tcp://*',
min_port=PORT_RANGE[0],
max_port=PORT_RANGE[1],
max_tries=500)
class OBCIClient(object):
default_timeout = 5000
def __init__(self, server_addresses, zmq_context=None):
self.ctx = zmq_context if zmq_context else zmq.Context()
self.server_addresses = server_addresses
self.server_req_socket = None
self.init_server_socket(server_addresses)
# self = zmq.Poller()
# self.register(self.server_req_socket, zmq.POLLIN)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.dns = net.DNS()
def init_server_socket(self, srv_addrs):
if self.server_req_socket is not None:
print("server socket restart")
self.server_req_socket.close()
self.server_req_socket = self.ctx.socket(zmq.REQ)
for addr in srv_addrs:
print(addr)
self.server_req_socket.connect(addr)
def launch(self, launch_file=None, sandbox_dir=None, name=None, overwrites=None):
result = self.send_create_experiment(launch_file, sandbox_dir, name, overwrites)
print("create result:", result)
if not result:
self.init_server_socket(self.server_addresses)
return result
if result.type != "experiment_created":
return result
print(result)
machine = result.origin_machine
addrs = [addr for addr in result.rep_addrs if self._addr_connectable(addr, machine)]
return self.send_start_experiment(addrs)
def morph(self, exp_strname, launch_file, name=None, overwrites=None, leave_on=None):
response = self.get_experiment_contact(exp_strname)
exp_sock = self.ctx.socket(zmq.REQ)
try:
if response.type == "rq_error" or response.type == "no_data":
return response
for addr in response.rep_addrs:
exp_sock.connect(addr)
msg = self.mtool.fill_msg('morph_to_new_scenario', launch_file=launch_file,
name=name, overwrites=overwrites, leave_on=leave_on)
send_msg(exp_sock, msg)
response, details = self.poll_recv(exp_sock, 6000)
return response
finally:
exp_sock.close()
def _addr_connectable(self, addr, machine):
return machine == socket.gethostname() or \
(net.is_ip(addr) and not net.addr_is_local(addr))
def start_chosen_experiment(self, exp_strname):
response = self.get_experiment_contact(exp_strname)
if response.type == "rq_error" or response.type == "no_data":
return response
return self.send_start_experiment(response.rep_addrs)
def send_start_experiment(self, exp_addrs):
exp_sock = self.ctx.socket(zmq.REQ)
try:
for addr in exp_addrs:
exp_sock.connect(addr)
send_msg(exp_sock, self.mtool.fill_msg("start_experiment"))
reply, details = self.poll_recv(exp_sock, 20000)
# print reply
return reply
finally:
exp_sock.close()
def force_kill_experiment(self, strname):
pass
def get_experiment_contact(self, strname):
send_msg(self.server_req_socket, self.mtool.fill_msg("get_experiment_contact",
strname=strname))
response, details = self.poll_recv(self.server_req_socket, self.default_timeout)
return response
def ping_server(self, timeout=50):
send_msg(self.server_req_socket, self.mtool.fill_msg("ping"))
response, details = self.poll_recv(self.server_req_socket, timeout)
return response
def retry_ping(self, timeout=50):
response, details = self.poll_recv(self.server_req_socket, timeout)
return response
def send_create_experiment(self, launch_file=None, sandbox_dir=None, name=None, overwrites=None):
send_msg(
self.server_req_socket,
self.mtool.fill_msg(
"create_experiment",
launch_file=launch_file,
sandbox_dir=sandbox_dir,
name=name,
overwrites=overwrites))
response, details = self.poll_recv(self.server_req_socket, 5000)
return response
def send_list_experiments(self):
send_msg(self.server_req_socket, self.mtool.fill_msg("list_experiments"))
response, details = self.poll_recv(self.server_req_socket, 4000)
return response
def send_list_nearby_machines(self):
send_msg(self.server_req_socket, self.mtool.fill_msg("list_nearby_machines"))
response, details = self.poll_recv(self.server_req_socket, 4000)
return response
def get_experiment_details(self, strname, peer_id=None):
response = self.get_experiment_contact(strname)
if response.type == "rq_error" or response.type == "no_data":
return response
sock = self.ctx.socket(zmq.REQ)
try:
for addr in response.rep_addrs:
sock.connect(addr)
if peer_id:
send_msg(sock, self.mtool.fill_msg("get_peer_info", peer_id=peer_id))
else:
send_msg(sock, self.mtool.fill_msg("get_experiment_info"))
response, details = self.poll_recv(sock, 2000)
return response
finally:
sock.close()
def configure_peer(self, exp_strname, peer_id, config_overrides, override_files=None):
response = self.get_experiment_contact(exp_strname)
if response.type == "rq_error" or response.type == "no_data":
return response
sock = self.ctx.socket(zmq.REQ)
try:
for addr in response.rep_addrs:
sock.connect(addr)
if override_files:
send_msg(sock, self.mtool.fill_msg("get_peer_info", peer_id=peer_id))
response, details = self.poll_recv(sock, 2000)
if response.type is 'rq_error':
return response
msg = self.mtool.fill_msg("update_peer_config", peer_id=peer_id,
**config_overrides)
send_msg(sock, msg)
# print msg
response, details = self.poll_recv(sock, 2000)
return response
finally:
sock.close()
# "update_peer_config" : dict(peer_id='', local_params='',
# external_params='', launch_dependencies='', config_sources=''),
def kill_exp(self, strname, force=False):
send_msg(self.server_req_socket,
self.mtool.fill_msg("kill_experiment", strname=strname))
return self.poll_recv(self.server_req_socket, 2000)[0]
def srv_kill(self):
send_msg(self.server_req_socket,
self.mtool.fill_msg("kill"))
return self.poll_recv(self.server_req_socket, 2000)[0]
def join_experiment(self, strname, peer_id, path):
response = self.get_experiment_contact(strname)
if response.type == "rq_error" or response.type == "no_data":
return response
sock = self.ctx.socket(zmq.REQ)
try:
self._connect(sock, response.rep_addrs)
send_msg(sock, self.mtool.fill_msg("join_experiment",
peer_id=peer_id, path=path, peer_type='obci_peer'))
response, details = self.poll_recv(sock, 5000)
sock.close()
return response
finally:
sock.close()
def add_peer(self, strname, peer_id, path, machine, param_overwrites=None,
custom_config_path=None, config_sources=None, launch_dependencies=None,
apply_globals=True):
response = self.get_experiment_contact(strname)
if response.type == "rq_error" or response.type == "no_data":
return response
sock = self.ctx.socket(zmq.REQ)
try:
self._connect(sock, response.rep_addrs)
send_msg(sock, self.mtool.fill_msg("add_peer",
peer_id=peer_id, peer_path=path, peer_type='obci_peer',
machine=machine, param_overwrites=param_overwrites,
custom_config_path=custom_config_path, config_sources=config_sources,
launch_dependencies=launch_dependencies, apply_globals=apply_globals))
response, details = self.poll_recv(sock, 5000)
return response
finally:
sock.close()
def kill_peer(self, exp_strname, peer_id, remove_config=False):
response = self.get_experiment_contact(exp_strname)
if response.type == "rq_error" or response.type == "no_data":
return response
try:
sock = self.ctx.socket(zmq.REQ)
self._connect(sock, response.rep_addrs)
send_msg(sock, self.mtool.fill_msg("kill_peer",
peer_id=peer_id, remove_config=remove_config))
response, details = self.poll_recv(sock, 5000)
return response
finally:
sock.close()
def _connect(self, sock, addr_list):
print("**** ", addr_list)
this = self._is_this_machine(addr_list)
connected = False
for addr in addr_list:
if not this and ('localhost' in addr or '127.0.0.1' in addr):
continue
try:
sock.connect(addr)
connected = True
except zmq.ZMQError as e:
print(addr, " ::: ", str(e))
if not connected:
raise Exception("Could not connect to any of the addresses: " + str(addr_list))
def _is_this_machine(self, addr_list):
for addr in addr_list:
if self.dns.is_this_machine(addr):
return True
return False
def leave_experiment(self, strname, peer_id):
response = self.get_experiment_contact(strname)
if response.type == "rq_error" or response.type == "no_data":
return response
try:
sock = self.ctx.socket(zmq.REQ)
self._connect(sock, response.rep_addrs)
send_msg(sock, self.mtool.fill_msg("leave_experiment",
peer_id=peer_id))
response, details = self.poll_recv(sock, 5000)
return response
finally:
sock.close()
def get_tail(self, strname, peer_id, len_):
response = self.get_experiment_contact(strname)
if response.type == "rq_error" or response.type == "no_data":
return response
sock = self.ctx.socket(zmq.REQ)
try:
for addr in response.rep_addrs:
sock.connect(addr)
send_msg(sock, self.mtool.fill_msg("get_tail", peer_id=peer_id, len=len_))
response, details = self.poll_recv(sock, 4000)
return response
finally:
sock.close()
def poll_recv(self, socket, timeout):
result, details = self.poller.poll_recv(socket, timeout)
if result:
result = self.mtool.unpack_msg(result)
else:
result = EmptyResponse(details)
return result, details
class SubprocessMonitor(object):
def __init__(self, zmq_ctx, uuid, logger=None, obci_dns=None):
self._processes = {}
self._ctx = zmq_ctx
self.uuid = uuid
self.logger = logger or get_logger('subprocess_monitor',
stream_level='warning')
self.obci_dns = obci_dns
self._mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self._proc_lock = threading.RLock()
def not_running_processes(self):
status = {}
with self._proc_lock:
for key, proc in self._processes.iteritems():
st = proc.status()
if st[0] in [FINISHED, FAILED, TERMINATED
] and not proc.marked_delete():
status[key] = st
return status
def unknown_status_processes(self):
with self._proc_lock:
return [
proc for proc in self._processes.values()
if proc.status()[0] == UNKNOWN
]
def process(self, machine_ip, pid):
with self._proc_lock:
return self._processes.get((machine_ip, pid), None)
def killall(self, force=False):
with self._proc_lock:
for proc in self._processes.values():
kill_method = proc.kill if not force else proc.kill_with_force
if proc.status()[0] not in [FINISHED, FAILED, TERMINATED]:
kill_method()
def delete(self, machine, pid):
proc = self._processes.get((machine, pid), None)
if proc is None:
raise Exception("Process not found: " + str((machine, pid)))
if not proc.running():
del self._processes[(machine, pid)]
return True
else:
self.logger.error("Process is running, will not delete! " +
str((machine, pid)))
return False
def delete_all(self):
with self._proc_lock:
for proc in self._processes.values():
del proc
self._processes = {}
def stop_monitoring(self):
with self._proc_lock:
for proc in self._processes.values():
proc.stop_monitoring()
def _launch_args(self, path, args):
# TODO fix interpreter calls // only python is supported
if path.endswith('.py'):
launch_args = PYTHON_CALL + [path] + args
else:
launch_args = [path] + args
return launch_args
def _stdio_actions(self, io_flags):
out = subprocess.PIPE if io_flags & STDOUT else None
if io_flags & STDERR:
err = subprocess.PIPE
elif out is not None:
err = subprocess.STDOUT
else:
err = None
stdin = subprocess.PIPE if io_flags & STDIN else None
return (out, err, stdin)
def _local_launch(self, launch_args, stdio_actions, env):
ON_POSIX = 'posix' in sys.builtin_module_names
out, err, stdin = stdio_actions
try:
if sys.platform == "win32":
crflags = subprocess.CREATE_NEW_PROCESS_GROUP
popen_obj = subprocess.Popen(launch_args,
stdout=out,
stderr=err,
stdin=stdin,
bufsize=1,
close_fds=ON_POSIX,
env=env,
creationflags=crflags)
else:
popen_obj = subprocess.Popen(launch_args,
stdout=out,
stderr=err,
stdin=stdin,
bufsize=1,
close_fds=ON_POSIX,
env=env)
details = "Popen constructor finished for " +\
str(launch_args[:3]) + "(...)"
self.logger.info(details)
return popen_obj, details
except OSError as e:
details = "Unable to spawn process {0} [{1}]".format(
launch_args, e.args)
self.logger.error(details)
return None, details
except ValueError as e:
details = "Unable to spawn process (bad arguments) {0} [{1}]".format(
launch_args, e.args)
self.logger.error(details)
return None, details
except Exception as e:
details = "Process launch Error: " + str(e) + str(e.args) + str(
vars(e))
self.logger.error(details)
return None, details
def new_local_process(self,
path,
args,
proc_type='',
name='',
capture_io=STDOUT | STDIN,
stdout_log=None,
stderr_log=None,
register_timeout_desc=None,
monitoring_optflags=RETURNCODE | PING,
machine_ip=None,
env=None):
launch_args = self._launch_args(path, args)
self.logger.debug(proc_type + " local path: " + path)
machine = machine_ip if machine_ip else socket.gethostname()
std_actions = self._stdio_actions(capture_io)
timeout_desc = register_timeout_desc
self.logger.debug('process launch arg list: %s', launch_args)
popen_obj, details = self._local_launch(launch_args, std_actions, env)
if popen_obj is None:
return None, details
if popen_obj.returncode is not None:
det = "opened process already terminated" + popen_obj.communicate()
self.logger.warning(det)
if not name:
name = os.path.basename(path)
process_desc = ProcessDescription(proc_type=proc_type,
name=name,
path=path,
args=args,
machine_ip=machine,
pid=popen_obj.pid)
# io_handler will be None if no stdio is captured
io_handler = start_stdio_handler(popen_obj, std_actions,
':'.join([machine, path, name]),
stdout_log, stderr_log)
new_proc = LocalProcess(process_desc,
popen_obj,
io_handler=io_handler,
reg_timeout_desc=timeout_desc,
monitoring_optflags=monitoring_optflags,
logger=self.logger)
if monitoring_optflags & PING:
new_proc._ctx = self._ctx
with self._proc_lock:
self._processes[(machine, popen_obj.pid)] = new_proc
new_proc.start_monitoring()
return new_proc, None
def new_remote_process(self,
path,
args,
proc_type,
name,
machine_ip,
conn_addr,
capture_io=STDOUT | STDIN,
stdout_log=None,
stderr_log=None,
register_timeout_desc=None,
monitoring_optflags=PING):
"""Send a request to conn_addr for a process launch. By default
the process will be monitored with ping requests and locally by the
remote peer."""
timeout_desc = register_timeout_desc
rq_message = self._mtool.fill_msg('launch_process',
path=path,
args=args,
proc_type=proc_type,
name=name,
machine_ip=machine_ip,
capture_io=capture_io,
stdout_log=stdout_log,
stderr_log=stderr_log)
rq_sock = self._ctx.socket(zmq.REQ)
try:
rq_sock.connect(conn_addr)
except zmq.ZMQError as e:
det = "Could not connect to {0}, err: {1}, {2}".format(
conn_addr, e, e.args)
self.logger.error(det)
return None, det
self.logger.info("SENDING LAUNCH REQUEST {0} {1} {2} {3}".format(
machine_ip, _DEFAULT_TIMEOUT_MS, 'ms', conn_addr))
send_msg(rq_sock, rq_message)
result, details = self.poller.poll_recv(rq_sock, _DEFAULT_TIMEOUT_MS)
rq_sock.close()
if not result:
details = details + " [address was: {0}]".format(conn_addr)
self.logger.error(details)
return None, details
else:
result = self._mtool.unpack_msg(result)
if result.type == 'rq_error':
det = "REQUEST FAILED" + str(result.err_code) + ':' + str(
result.details)
self.logger.error(det)
return None, det
elif result.type == 'launched_process_info':
self.logger.info("REQUEST SUCCESS %s", result.dict())
process_desc = ProcessDescription(proc_type=result.proc_type,
name=result.name,
path=result.path,
args=args,
machine_ip=result.machine,
pid=result.pid)
new_proc = RemoteProcess(process_desc,
conn_addr,
reg_timeout_desc=timeout_desc,
monitoring_optflags=monitoring_optflags,
logger=self.logger)
if monitoring_optflags & PING:
new_proc._ctx = self._ctx
with self._proc_lock:
self._processes[(result.machine, result.pid)] = new_proc
new_proc.start_monitoring()
return new_proc, None
class Process(object):
def __init__(self, proc_description,
reg_timeout_desc=None,
monitoring_optflags=PING,
logger=None):
self.desc = proc_description
self.must_register = reg_timeout_desc is not None
self._status_lock = threading.RLock()
self._status = UNKNOWN if self.must_register else RUNNING
self._status_details = None
self.ping_it = monitoring_optflags & PING
self.check_returncode = monitoring_optflags & RETURNCODE if \
self.desc.pid is not None else False
self.logger = logger or get_logger(
'subprocess_monitor'+'-'+self.desc.name+'-'+str(self.desc.pid),
stream_level='info')
self.set_registration_timeout_handler(reg_timeout_desc)
self.registration_data = None
self._stop_monitoring = False
self._ping_thread = None
self._ping_retries = 8
self._returncode_thread = None
self._mtool = OBCIMessageTool(message_templates)
self._ctx = None
self.rq_sock = None
self._poller = PollingObject()
self.delete = False
@property
def machine_ip(self):
return self.desc.machine_ip
@property
def pid(self):
return self.desc.pid
@property
def path(self):
return self.desc.path
@property
def proc_type(self):
return self.desc.proc_type
@property
def name(self):
return self.desc.name
def status(self):
with self._status_lock:
return self._status, self._status_details
def set_registration_timeout_handler(self, reg_timeout_desc):
with self._status_lock:
self._status = UNKNOWN
self._status_details = None
self.must_register = reg_timeout_desc is not None
self.reg_timeout_desc = reg_timeout_desc
self.reg_timer = None if not self.must_register else \
self.new_timer(self.reg_timeout_desc, REG_TIMER)
if self.must_register:
self.reg_timer.start()
def is_local(self):
raise NotImplementedError()
def timeout_handler(self, custom_method, args, type_):
self._do_handle_timeout(type_)
custom_method(*args)
def _do_handle_timeout(self, type_):
raise NotImplementedError()
def new_timer(self, tim_desc, type_):
return threading.Timer(tim_desc.timeout, self.timeout_handler,
[tim_desc.timeout_method, tim_desc.timeout_args, type_])
def registered(self, reg_data):
if self.reg_timer is not None:
self.reg_timer.cancel()
self.logger.info("{0} [{1}] REGISTERED!!! {2}".format(
self.name, self.proc_type, reg_data.machine_ip))
#print "ping:", self.ping_it, "ret:", self.check_returncode
with self._status_lock:
self._status = RUNNING
#TODO validate registration data
self.registration_data = reg_data
self.logger.info("reg_data" + str(vars(reg_data)))
if self.ping_it:
if not self._ctx:
self._ctx = zmq.Context()
self.rq_sock = self._ctx.socket(zmq.REQ)
for addr in reg_data.rep_addrs:
if reg_data.machine_ip != socket.gethostname() and\
net.addr_is_local(addr):
continue
self.logger.debug(self.name + "connecting to " + addr)
self.rq_sock.connect(addr)
def stop_monitoring(self):
if self.reg_timer:
self.reg_timer.cancel()
self.reg_timer = None
self._stop_monitoring = True
if self._ping_thread is not None:
self.logger.info("%s, %s, %s",
self.proc_type, self.name ,"Joining ping thread")
self._ping_thread.join()
if self._returncode_thread is not None:
self.logger.info("%s %s %s",
self.proc_type,self.name, "joining returncode thread")
self._returncode_thread.join()
self.logger.info("monitor for: %s, %s, %s",
self.proc_type,self.name, " ...monitoring threads stopped.")
def finished(self):
finished = True
if self._ping_thread is not None:
finished = not self._ping_thread.is_alive()
if self._returncode_thread is not None:
finished = finished and not self._returncode_thread.is_alive()
return finished
def process_is_running(self):
running = True
if self._ping_thread is not None:
running = self._ping_thread.is_alive()
if self._returncode_thread is not None:
running = running and self._returncode_thread.is_alive()
return running
def start_monitoring(self):
if self.ping_it:
self._ping_thread = threading.Thread(target=self.ping_monitor, args=())
self._ping_thread.daemon = True
self._ping_thread.start()
if self.check_returncode:
self._returncode_thread = threading.Thread(target=self.returncode_monitor, args=())
self._returncode_thread.daemon = True
self._returncode_thread.start()
def ping_monitor(self):
is_alive = True
try:
while not self._stop_monitoring and is_alive:
time.sleep(2)
if self.rq_sock is not None:
send_msg(self.rq_sock, self._mtool.fill_msg('ping'))
result = None
while self._ping_retries and not result and not self._stop_monitoring:
result, det = self._poller.poll_recv(socket=self.rq_sock, timeout=1500)
if not result and not self._stop_monitoring:
self.logger.info("%s %s %s",
self.proc_type, self.name, "NO RESPONSE TO PING!")
with self._status_lock:
if self._status not in [FAILED, FINISHED]:
self._status = NON_RESPONSIVE
self._status_details = 'ping response timeout'
print "status:", self._status
is_alive = False
finally:
self.rq_sock.close(linger=0)
def returncode_monitor(self):
raise NotImplementedError()
def kill(self):
raise NotImplementedError()
def mark_delete(self):
with self._status_lock:
self.delete = True
def marked_delete(self):
with self._status_lock:
return self.delete
class EEGExperimentFinder(object):
def __init__(self, srv_addrs, ctx, client_push_address, nearby_servers):
self.ctx = ctx
self.server_req_socket = self.ctx.socket(zmq.REQ)
for addr in srv_addrs:
self.server_req_socket.connect(addr)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.nearby_servers = nearby_servers
self._amplified_cache = {}
def _running_experiments(self):
send_msg(self.server_req_socket, self.mtool.fill_msg("list_experiments"))
exp_list, details = self.poll_recv(self.server_req_socket, 2000)
if not exp_list:
LOGGER.error("Connection to obci_server failed. (list_experiments)")
return None
exps = exp_list.exp_data
running = []
for exp in exps.values():
if exp['status_name'] == launcher_tools.RUNNING or \
exp['status_name'] == launcher_tools.LAUNCHING:
running.append(exp)
return running
def find_amplified_experiments(self):
running_exps = self._running_experiments()
amplified = []
for exp in running_exps:
LOGGER.info("Found running experiment: " + str(exp['name']))
infos = self._info_amplified(exp)
if infos is not None:
print("Found experiments...", str(infos)[:500])
amplified += infos
return amplified
def _info_amplified(self, exp_desc):
amp_options = []
LOGGER.info("Processing experiment " + str(exp_desc['name']) +
"w/ addr: " + str(exp_desc['rep_addrs']))
tcp_addrs = exp_desc['tcp_addrs']
rep_addrs = net.choose_not_local(exp_desc['rep_addrs'])
if not rep_addrs:
rep_addrs = net.choose_local(exp_desc['rep_addrs'], ip=True)
rep_addr = rep_addrs.pop()
pub_addrs = net.choose_not_local(exp_desc['pub_addrs'])
if not pub_addrs:
pub_addrs = net.choose_local(exp_desc['pub_addrs'], ip=True)
pub_addr = pub_addrs.pop()
tcp_addr = tcp_addrs.pop()
LOGGER.info("Chosen experiment addresses: REP -- " +
str(rep_addr) + ", PUB -- " + str(pub_addr))
req_sock = self.ctx.socket(zmq.REQ)
try:
req_sock.connect(rep_addr)
send_msg(req_sock, self.mtool.fill_msg('get_experiment_info'))
res, details = self.poll_recv(req_sock, 4000)
finally:
req_sock.close()
if not res:
LOGGER.error("Connection failed (experiment " + exp_desc['name'] +
"), get_experiment_info")
return None
exp_info = res.dict() # json.loads(res)
for field in ["peers_status", "details"]:
del exp_info["experiment_status"][field]
peer_list = exp_info["peers"]
if not self._has_mx(peer_list):
LOGGER.info("Experiment " + exp_desc['name'] +
" does not have a multiplexer.")
return None
maybe_amps = self._amp_like_peers(peer_list)
if not maybe_amps:
LOGGER.info("Experiment " + exp_desc['name'] +
" -- no amplifier.")
return None
req_sock = self.ctx.socket(zmq.REQ)
try:
req_sock.connect(rep_addr)
for peer in maybe_amps:
info, params = self._get_amp_info(req_sock, peer)
if not self._is_amplifier(info, params):
LOGGER.info("Experiment " + exp_desc['name'] +
" -- peer " + str(peer) + "is not an amplifier.")
continue
else:
exp_data = self._create_exp_data(exp_info, info, params['param_values'],
rep_addr, pub_addr, tcp_addr)
amp_options.append(exp_data)
finally:
req_sock.close()
return amp_options
def _get_amp_info(self, exp_sock, peer_id):
send_msg(exp_sock, self.mtool.fill_msg('get_peer_info', peer_id=peer_id))
info, details = self.poll_recv(exp_sock, 4000)
send_msg(exp_sock, self.mtool.fill_msg('get_peer_param_values', peer_id=peer_id))
params, details = self.poll_recv(exp_sock, 4000)
if not info or not params:
LOGGER.error("get_peer_info failed " + str(peer_id) + " " + str(details))
return None, None
info = info.dict()
params = params.dict()
for field in ["sender", "sender_ip", "receiver", "type", "local_params", "external_params",
"config_sources", "launch_dependencies"]:
del info[field]
return info, params
def _is_amplifier(self, peer_info, peer_params):
info = peer_info
peer_id = info['peer_id']
if not info['peer_type'] == 'obci_peer':
LOGGER.info("Peer " + str(peer_id) + " not obci_peer")
return False
params = peer_params['param_values']
if 'channels_info' not in params or\
'active_channels' not in params:
LOGGER.info('Peer ' + str(peer_id) + " no channels_info param.")
return False
return True
def _create_exp_data(self, exp_info, peer_info, params, rep_addr, pub_addr, tcp_addr):
data = {}
data['amplifier_params'] = params
data['amplifier_peer_info'] = peer_info
data['experiment_info'] = exp_info
data['rep_addrs'] = [rep_addr]
data['pub_addrs'] = [pub_addr]
data['tcp_addrs'] = [tcp_addr]
return data
def _has_mx(self, peer_list):
return [peer for peer in peer_list if peer.startswith('mx')] != []
def _amp_like_peers(self, peer_list):
return [peer for peer in peer_list if peer.startswith('amplifier')]
def poll_recv(self, socket, timeout):
result, details = self.poller.poll_recv(socket, timeout)
if result:
result = self.mtool.unpack_msg(result)
return result, details
class ExperimentEngineInfo(QtCore.QObject):
exp_saver_msg = QtCore.pyqtSignal(object)
def __init__(self, preset_data=None, launcher_data=None, ctx=None):
self.preset_data = preset_data
self.launch_file = None
self.launcher_data = launcher_data
self.name = None
self.info = ""
self.public_params = []
self.origin_machine = ''
self.unsupervised_peers = {}
self.old_uid = None
self.ctx = ctx
self.exp_req = None
self.mtool = OBCIMessageTool(message_templates)
self.poller = PollingObject()
self.category = DEFAULT_CATEGORY
if preset_data is not None:
self.setup_from_preset(preset_data)
elif launcher_data is not None:
self.setup_from_launcher(launcher_data)
super(ExperimentEngineInfo, self).__init__()
def cleanup(self):
if self.exp_req:
self.exp_req.close() # linger=0)
def setup_from_preset(self, preset_data, launcher=False):
self.preset_data = preset_data
self.overwrites = {}
self.runtime_changes = {}
self.status = launcher_tools.ExperimentStatus()
self.exp_config = system_config.OBCIExperimentConfig()
self.name = preset_data['name']
self.launch_file = preset_data['launch_file']
self.info = preset_data['info']
self.public_params = [p.strip() for p in preset_data['public_params'].split(',')]
self.exp_config.uuid = self.name + '--' + self.launch_file
self.category = preset_data['category']
result, details = self._make_config()
self.status.details = details
self._set_public_params()
def _addr_connectable(self, addr, machine):
return machine == socket.gethostname() or \
(net.is_ip(addr) and not net.addr_is_local(addr))
def setup_from_launcher(self, launcher_data, preset=False, transform=False):
self.launcher_data = launcher_data
self.runtime_changes = {}
if preset:
self.old_uid = self.exp_config.uuid
if not preset or transform:
self.overwrites = {}
self.status = launcher_tools.ExperimentStatus()
self.exp_config = system_config.OBCIExperimentConfig()
self.name = launcher_data['name'] # if not preset else self.old_uid
if not preset:
self.launch_file = launcher_data['launch_file_path']
connected = False
if not transform:
self.ctx = self.ctx if self.ctx is not None else zmq.Context()
self.exp_req = self.ctx.socket(zmq.REQ)
machine = launcher_data['origin_machine']
for addr in launcher_data['rep_addrs']:
if self._addr_connectable(addr, machine):
try:
self.exp_req.connect(addr)
except Exception as e:
print(addr, False)
else:
connected = True
if not connected:
print("Connection to experiment ", self.name, "UNSUCCESFUL!!!!!!")
return
self.exp_config.uuid = launcher_data['uuid']
self.exp_config.origin_machine = launcher_data['origin_machine']
self.uuid = self.exp_config.uuid
self.exp_config.launch_file_path = self.launch_file
result, details = self._get_experiment_scenario()
self.exp_config.status(self.status)
self.status.set_status(launcher_data['status_name'], details=launcher_data['details'])
self._get_experiment_details()
self._set_public_params()
def update_scenario(self, launch_file_path, scenario):
self.exp_config.launch_file_path = launch_file_path
self._process_experiment_scenario(scenario)
self.exp_config.status(self.status)
def _set_public_params(self):
for par in self.public_params:
if len(par.split('.')) == 2:
[peer, param] = par.split('.')
self.exp_config.peers[peer].public_params.append(param)
def _make_config(self):
self.exp_config.launch_file_path = self.launch_file
self.uuid = self.exp_config.uuid
result, details = self.make_experiment_config()
self.exp_config.status(self.status)
return result, details
# FIXME !!! copy-paste from obci_experiment
def make_experiment_config(self):
launch_parser = launch_file_parser.LaunchFileParser(
launcher_tools.obci_root(), settings.DEFAULT_SCENARIO_DIR)
if not self.launch_file:
return False, "Empty scenario."
try:
with open(launcher_tools.expand_path(self.launch_file)) as f:
launch_parser.parse(f, self.exp_config, apply_globals=True)
except Exception as e:
self.status.set_status(launcher_tools.NOT_READY, details=str(e))
print("config errror ", str(e))
return False, str(e)
rd, details = self.exp_config.config_ready()
if rd:
self.status.set_status(launcher_tools.READY_TO_LAUNCH)
else:
self.status.set_status(launcher_tools.NOT_READY, details=details)
print(rd, details)
return True, None
def _get_experiment_scenario(self):
if not self.exp_req:
return False, "No experiment socket"
response = self.comm_exp(self.mtool.fill_msg("get_experiment_scenario"))
if not response:
return False, "No response from experient"
print("GOT SCENARIO", response.scenario)
return self._process_experiment_scenario(response.scenario)
def _process_experiment_scenario(self, json_scenario):
jsonpar = launch_file_parser.LaunchJSONParser(
launcher_tools.obci_root(), settings.DEFAULT_SCENARIO_DIR)
inbuf = io.StringIO(json_scenario)
jsonpar.parse(inbuf, self.exp_config)
print("MY PEEEEERS:", self.exp_config.peers.keys())
rd, details = self.exp_config.config_ready()
if rd:
self.status.set_status(launcher_tools.READY_TO_LAUNCH)
else:
self.status.set_status(launcher_tools.NOT_READY, details=details)
print(rd, details)
return True, None
def _get_experiment_details(self):
if not self.exp_req:
return
exp_msg = self.comm_exp(self.mtool.fill_msg("get_experiment_info"))
if not exp_msg:
return
self.origin_machine = exp_msg.origin_machine
for peer, short_info in exp_msg.peers.items():
# self.exp_config.set_peer_machine(peer, short_info['machine'])
msg = self.comm_exp(self.mtool.fill_msg("get_peer_info",
peer_id=peer))
if not msg:
return
ext_defs = {}
for name, defi in msg.external_params.items():
ext_defs[name] = defi[0] + '.' + defi[1]
self.exp_config.update_peer_config(peer, dict(config_sources=msg.config_sources,
launch_dependencies=msg.launch_dependencies,
local_params=msg.local_params,
external_params=ext_defs))
for peer, status in exp_msg.experiment_status['peers_status'].items():
self.status.peer_status(peer).set_status(
status['status_name'],
details=status['details'])
def parameters(self, peer_id, mode):
params = {}
peer = self.exp_config.peers[peer_id]
if mode == MODE_BASIC:
for par in peer.public_params:
params[par] = (self.exp_config.param_value(peer_id, par), None)
else:
params = peer.config.local_params
for param in peer.config.local_params:
params[param] = (self.exp_config.param_value(peer_id, param), None)
for param, defi in peer.config.ext_param_defs.items():
source_symbol = defi[0]
source = peer.config.config_sources[source_symbol]
params[param] = (self.exp_config.param_value(peer_id, param), source + '.' + defi[1])
return params
def comm_exp(self, msg):
send_msg(self.exp_req, msg)
response, _ = self.poller.poll_recv(self.exp_req, timeout=3000)
if not response:
print("!!!!!!!!!!!!!!!!!!!!!!!!!!!1 no response to ", msg)
self.exp_req.close()
self.exp_req = self.ctx.socket(zmq.REQ)
for addr in self.launcher_data['rep_addrs']:
if self._addr_connectable(addr, self.launcher_data['origin_machine']):
self.exp_req.connect(addr)
return None
return self.mtool.unpack_msg(response)
def updatable(self, peer_id, config_part, **kwargs):
return False
def update_peer_param(self, peer_id, param, value, runtime=False):
changes = self.overwrites if not runtime else self.runtime_changes
ovr = changes.get(peer_id, None)
ovr = ovr if ovr is not None else {}
if param not in ovr:
old = self.exp_config.param_value(peer_id, param)
if old != value:
ovr[param] = old
changes[peer_id] = ovr
self.exp_config.update_local_param(peer_id, param, value)
def get_launch_args(self):
d = dict(launch_file=self.launch_file, name=self.name)
args = ['--ovr']
if self.overwrites:
for peer_id in self.overwrites:
args.append('--peer')
args.append(peer_id)
for arg in self.overwrites[peer_id]:
args += ['-p', arg, self.exp_config.param_value(peer_id, arg)]
pack = peer_cmd.peer_overwrites_pack(args)
d['overwrites'] = pack
print("overwrites pack!!!!!!!!!!!!!!!!!!!!! ", pack)
return d
def peer_info(self, peer_id):
return self.exp_config.peers[peer_id]
def add_peer(self, peer_id, peer_path, config_sources=None, launch_deps=None,
custom_config_path=None, param_overwrites=None, machine=None):
return launch_file_parser.extend_experiment_config(
self.exp_config,
peer_id,
peer_path,
config_sources,
launch_deps,
custom_config_path,
param_overwrites,
machine,
apply_globals=True)
def enable_signal_storing(self, store_options):
if not store_options:
return
if int(store_options['append_timestamp']):
store_options = dict(store_options)
store_options['save_file_name'] = store_options['save_file_name'] + "_" + str(time.time())
for peer, peer_path in SIGNAL_STORAGE_PEERS.items():
if peer not in self.exp_config.peers:
self.add_peer(peer, peer_path)
saver = self.exp_config.peers['signal_saver']
params = saver.config.param_values
for opt, val in store_options.items():
if opt in saver.config.param_values:
params[opt] = val
def stop_storing(self, client):
join_response = client.join_experiment(self.uuid, "dummy_module_" + str(time.time()), "")
if join_response is None:
print("experiment engine info - ERROR - connection timeout on stop signal storing!")
return
if not join_response.type == "rq_ok":
print("experiment engine info - ERROR - join error on stop signal storing!")
return
mx_addr = join_response.params["mx_addr"].split(':')
# hang and wait ...
acquisition_helper.finish_saving([(mx_addr[0], int(mx_addr[1]))])
class EEGExperimentFinder(object):
def __init__(self, srv_addrs, ctx, client_push_address, nearby_servers):
self.ctx = ctx
self.server_req_socket = self.ctx.socket(zmq.REQ)
for addr in srv_addrs:
self.server_req_socket.connect(addr)
self.poller = PollingObject()
self.mtool = OBCIMessageTool(message_templates)
self.nearby_servers = nearby_servers
self._amplified_cache = {}
def _running_experiments(self):
send_msg(self.server_req_socket,
self.mtool.fill_msg("list_experiments"))
exp_list, details = self.poll_recv(self.server_req_socket, 2000)
if not exp_list:
LOGGER.error(
"Connection to obci_server failed. (list_experiments)")
return None
exps = exp_list.exp_data
running = []
for exp in exps.values():
if exp['status_name'] == launcher_tools.RUNNING or \
exp['status_name'] == launcher_tools.LAUNCHING:
running.append(exp)
return running
def find_amplified_experiments(self):
running_exps = self._running_experiments()
amplified = []
for exp in running_exps:
LOGGER.info("Found running experiment: " + str(exp['name']))
infos = self._info_amplified(exp)
if infos is not None:
print "Found experiments...", str(infos)[:500]
amplified += infos
return amplified
def _info_amplified(self, exp_desc):
amp_options = []
LOGGER.info("Processing experiment "+ str(exp_desc['name']) +\
"w/ addr: " + str(exp_desc['rep_addrs']))
tcp_addrs = exp_desc['tcp_addrs']
rep_addrs = net.choose_not_local(exp_desc['rep_addrs'])
if not rep_addrs:
rep_addrs = net.choose_local(exp_desc['rep_addrs'], ip=True)
rep_addr = rep_addrs.pop()
pub_addrs = net.choose_not_local(exp_desc['pub_addrs'])
if not pub_addrs:
pub_addrs = net.choose_local(exp_desc['pub_addrs'], ip=True)
pub_addr = pub_addrs.pop()
tcp_addr = tcp_addrs.pop()
LOGGER.info("Chosen experiment addresses: REP -- " + \
str(rep_addr) + ", PUB -- " + str(pub_addr))
req_sock = self.ctx.socket(zmq.REQ)
try:
req_sock.connect(rep_addr)
send_msg(req_sock, self.mtool.fill_msg('get_experiment_info'))
res, details = self.poll_recv(req_sock, 4000)
finally:
req_sock.close()
if not res:
LOGGER.error("Connection failed (experiment " + exp_desc['name'] + \
"), get_experiment_info")
return None
exp_info = res.dict() #json.loads(res)
for field in ["peers_status", "details"]:
del exp_info["experiment_status"][field]
peer_list = exp_info["peers"]
if not self._has_mx(peer_list):
LOGGER.info("Experiment " + exp_desc['name'] + \
" does not have a multiplexer.")
return None
maybe_amps = self._amp_like_peers(peer_list)
if not maybe_amps:
LOGGER.info("Experiment "+ exp_desc['name'] + \
" -- no amplifier.")
return None
req_sock = self.ctx.socket(zmq.REQ)
try:
req_sock.connect(rep_addr)
for peer in maybe_amps:
info, params = self._get_amp_info(req_sock, peer)
if not self._is_amplifier(info, params):
LOGGER.info("Experiment "+ exp_desc['name'] + \
" -- peer " + str(peer) + "is not an amplifier.")
continue
else:
exp_data = self._create_exp_data(exp_info, info,
params['param_values'],
rep_addr, pub_addr,
tcp_addr)
amp_options.append(exp_data)
finally:
req_sock.close()
return amp_options
def _get_amp_info(self, exp_sock, peer_id):
send_msg(exp_sock, self.mtool.fill_msg('get_peer_info',
peer_id=peer_id))
info, details = self.poll_recv(exp_sock, 4000)
send_msg(exp_sock,
self.mtool.fill_msg('get_peer_param_values', peer_id=peer_id))
params, details = self.poll_recv(exp_sock, 4000)
if not info or not params:
LOGGER.error("get_peer_info failed " + str(peer_id) + " " +
str(details))
return None, None
info = info.dict()
params = params.dict()
for field in [
"sender", "sender_ip", "receiver", "type", "local_params",
"external_params", "config_sources", "launch_dependencies"
]:
del info[field]
return info, params
def _is_amplifier(self, peer_info, peer_params):
info = peer_info
peer_id = info['peer_id']
if not info['peer_type'] == 'obci_peer':
LOGGER.info("Peer " + str(peer_id) + " not obci_peer")
return False
params = peer_params['param_values']
if not 'channels_info' in params or\
not 'active_channels' in params:
LOGGER.info('Peer ' + str(peer_id) + " no channels_info param.")
return False
return True
def _create_exp_data(self, exp_info, peer_info, params, rep_addr, pub_addr,
tcp_addr):
data = {}
data['amplifier_params'] = params
data['amplifier_peer_info'] = peer_info
data['experiment_info'] = exp_info
data['rep_addrs'] = [rep_addr]
data['pub_addrs'] = [pub_addr]
data['tcp_addrs'] = [tcp_addr]
return data
def _has_mx(self, peer_list):
return [peer for peer in peer_list if peer.startswith('mx')] != []
def _amp_like_peers(self, peer_list):
return [peer for peer in peer_list if peer.startswith('amplifier')]
def poll_recv(self, socket, timeout):
result, details = self.poller.poll_recv(socket, timeout)
if result:
result = self.mtool.unpack_msg(result)
return result, details