def _send2one(self,content,ip,port):
try:
s=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.info('connecting to %s:%d...'%(ip,port))
s.connect((ip,port))
s.send(content)
s.close()
LOG.info('connection to %s:%d close'%(ip,port))
except socket.timeout,e:
LOG.warning('%s:%d lost response'%(ip,port))
def _send2one(self, content, ip, port):
try:
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.info('connecting to %s:%d...' % (ip, port))
s.connect((ip, port))
s.send(content)
s.close()
LOG.info('connection to %s:%d close' % (ip, port))
except socket.timeout, e:
LOG.warning('%s:%d lost response' % (ip, port))
def _send2one(self,content,ip,port):
try:
s=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.debug('connect to %s:%d...'%(ip,port))
s.connect((ip,port))
s.send(content)
s.close()
LOG.debug('connection to %s:%d close'%(ip,port))
except socket.timeout as e:
LOG.warning('%s:%d lost response'%(ip,port))
except socket.error as arg:
LOG.error('socket error while connecting to %s:%d errno %d: %s'%(ip,port,arg[0],arg[1]))
def _send2one(self, content, ip, port):
try:
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.debug('connect to %s:%d...' % (ip, port))
s.connect((ip, port))
s.send(content)
s.close()
LOG.debug('connection to %s:%d close' % (ip, port))
except socket.timeout as e:
LOG.warning('%s:%d lost response' % (ip, port))
except socket.error as arg:
LOG.error('socket error while connecting to %s:%d errno %d: %s' %
(ip, port, arg[0], arg[1]))
def _send2one_r(self,content,ip,port,result):
try:
s=socket.socket(socket.AF_INET,socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.info('connecting to %s:%d...'%(ip,port))
s.connect((ip,port))
s.send(content.replace('__EOF__','__EOF___'))
s.send('__EOF__')
r=s.recv(4096)
if r!='':
result.append(r)
s.close()
LOG.debug('connection to %s:%d close'%(ip,port))
except socket.timeout,e:
LOG.warning('%s:%d lost response'%(ip,port))
return ''
def _send2one_r(self, content, ip, port, result):
try:
s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
s.settimeout(self._args.timeout)
LOG.info('connecting to %s:%d...' % (ip, port))
s.connect((ip, port))
s.send(content.replace('__EOF__', '__EOF___'))
s.send('__EOF__')
r = s.recv(4096)
if r != '':
result.append(r)
s.close()
LOG.debug('connection to %s:%d close' % (ip, port))
except socket.timeout, e:
LOG.warning('%s:%d lost response' % (ip, port))
return ''
def parse_distribute_task(self):
# do some common check here
if self._args.task_granularity < 0 or self._args.task_granularity > 3:
raise SwarmUseException(
'invalid task granularity, it should be one number of 1-3')
if self._args.process_num < 0:
raise SwarmUseException('process number can not be negative')
if self._args.thread_num <= 0:
raise SwarmUseException('thread number should be positive')
# connect to db server
LOG.info('try to connect to db server: %s:%d' %
(self._args.db_addr, self._args.db_port))
self._args.db, self._args.coll = init_db(self._args.db_addr,
self._args.db_port,
self._args.mod)
LOG.info('Connection to db server completed')
# start the manager
self._manager = MSwarmManager(self._args.timeout,
address=('', self._args.m_port),
authkey=self._args.authkey)
try:
module = importlib.import_module('modules.' + self._args.mod +
'.' + self._args.mod)
except ImportError as e:
raise SwarmModuleException('an error occured when load module:' +
self._args.mod)
LOG.info('load module: ' + self._args.mod)
LOG.info('begin to decompose task...')
mod_master = getattr(module, 'Master')(self._args)
# begin first round of tasks decomposition and distribution
roundn = 0
self._manager.init_task_statistics()
while True:
subtaskl = mod_master.generate_subtasks()
taskn = len(subtaskl)
if taskn == 0:
break
roundn += 1
LOG.log(REPORT, 'begin round %d' % roundn)
LOG.info('round %d: put task into queue...' % roundn)
for cur in subtaskl:
self._manager.put_task(self._args.mod, cur)
LOG.log(
REPORT, 'round %d: %d tasks have been put into queue' %
(roundn, taskn))
LOG.info('round %d: get result from swarm...' % roundn)
# get result
confirmedn = 0
self._manager.prepare_get_result()
while True:
try:
result = self._manager.get_result()
if result == '':
break
confirmedn += 1
LOG.log(
REPORT, 'round %d: %d/%d tasks have been completed' %
(roundn, confirmedn, taskn))
mod_master.handle_result(result)
except Queue.Empty as e:
# check number of slave host, if someone has lost response, reorganize tasks
# in queue.
LOG.info('try to detect swarm status')
r = self._send2swarm_r('ack')
if len(r) < self._swarm_num:
LOG.warning(
'%d of swarm has lost response. now total swarm:%d'
% (self._swarm_num - len(r), len(r)))
self._swarm_num = len(r)
# if no swarm left
if self._swarm_num == 0:
raise SwarmSlaveException(
'no swarm left. task failed')
LOG.log(REPORT, 'reorganize tasks in queue...')
self._manager.reorganize_tasks()
LOG.log(REPORT, 'reorganization completed')
else:
LOG.log(
REPORT, 'all swarm works fine. now num: %d' %
self._swarm_num)
# continue
LOG.log(REPORT, 'round %d over' % roundn)
LOG.log(REPORT, 'all tasks have been comfirmed')
# do final report now
mod_master.report()
self._shutdown()
def parse_distribute_task(self):
# do some common check here
if self._args.task_granularity<0 or self._args.task_granularity>3:
raise SwarmUseException('invalid task granularity, it should be one number of 1-3')
if self._args.process_num<0:
raise SwarmUseException('process number can not be negative')
if self._args.thread_num<=0:
raise SwarmUseException('thread number should be positive')
# connect to db server
LOG.info('try to connect to db server: %s:%d'%(self._args.db_addr,self._args.db_port))
self._args.db,self._args.coll=init_db(self._args.db_addr,self._args.db_port,self._args.mod)
LOG.info('Connection to db server completed')
# start the manager
self._manager=MSwarmManager(self._args.timeout,address=('', self._args.m_port),
authkey=self._args.authkey)
try:
module=importlib.import_module('modules.'+self._args.mod+'.'+self._args.mod)
except ImportError as e:
raise SwarmModuleException('an error occured when load module:'+self._args.mod)
LOG.info('load module: '+self._args.mod)
LOG.info('begin to decompose task...')
mod_master=getattr(module,'Master')(self._args)
# begin first round of tasks decomposition and distribution
roundn=0
self._manager.init_task_statistics()
while True:
subtaskl=mod_master.generate_subtasks()
taskn=len(subtaskl)
if taskn==0:
break
roundn+=1
LOG.log(REPORT,'begin round %d'%roundn)
LOG.info('round %d: put task into queue...'%roundn)
for cur in subtaskl:
self._manager.put_task(self._args.mod,cur)
LOG.log(REPORT,'round %d: %d tasks have been put into queue'%(roundn,taskn))
LOG.info('round %d: get result from swarm...'%roundn)
# get result
confirmedn=0
self._manager.prepare_get_result()
while True:
try:
result=self._manager.get_result()
if result=='':
break
confirmedn+=1
LOG.log(REPORT,'round %d: %d/%d tasks have been completed'%(roundn,
confirmedn,taskn))
mod_master.handle_result(result)
except Queue.Empty as e:
# check number of slave host, if someone has lost response, reorganize tasks
# in queue.
LOG.info('try to detect swarm status')
r=self._send2swarm_r('ack')
if len(r)<self._swarm_num:
LOG.warning('%d of swarm has lost response. now total swarm:%d'
%(self._swarm_num-len(r),len(r)))
self._swarm_num=len(r)
# if no swarm left
if self._swarm_num==0:
raise SwarmSlaveException('no swarm left. task failed')
LOG.log(REPORT,'reorganize tasks in queue...')
self._manager.reorganize_tasks()
LOG.log(REPORT,'reorganization completed')
else:
LOG.log(REPORT,'all swarm works fine. now num: %d'%self._swarm_num)
# continue
LOG.log(REPORT,'round %d over'%roundn)
LOG.log(REPORT,'all tasks have been comfirmed')
# do final report now
mod_master.report()
self._shutdown()
