class InputThreadNew(threading.Thread):
def __init__(self, conf, processor=None, proc_name=None):
threading.Thread.__init__(self)
self.running = True
self.proc_name = proc_name # Only for logging
self.input_tube = conf['beanstalk_conf']['input_tube']
self.beanstalk = PyBeanstalk(conf['beanstalk_conf']['host'],
conf['beanstalk_conf']['port'])
self.out_beanstalk = PyBeanstalk(conf['beanstalk_conf']['host'],
conf['beanstalk_conf']['port'])
self.output_tube = conf['beanstalk_conf']['output_tube']
self.topic_output_tubes = {}
self.topic_output_tubes.setdefault('default', [])
"""
output_tube = ["default_out", "only_special_out:1,2,3:exclusive", "special_out:4", ":5:exclusive"]
topic_id:1,2,3只会用到only_special_out
topic_id:4 会进入special_out和default_out
topic_id:5 不会进入队列
topic_id:else 用用default_out队列
"""
if type(self.output_tube) == list:
for tube_def in self.output_tube:
tube_def = tube_def.strip()
if len(tube_def.split(":")) < 2:
self.topic_output_tubes['default'].append(tube_def)
else:
elements = [a.strip() for a in tube_def.split(':')]
tube_name = elements[0]
topic_ids = [
int(a.strip()) for a in elements[1].split(',')
]
exclusive = False
if len(elements) == 3 and elements[2] == 'exclusive':
exclusive = True
for topic_id in topic_ids:
self.topic_output_tubes.setdefault(topic_id, [])
self.topic_output_tubes[topic_id].append(
(tube_name, exclusive))
else:
self.topic_output_tubes['default'].append(self.output_tube)
self.log = log
if processor is None:
log.error("Processor not given !")
raise Exception("Processor not given !")
else:
self.processor = processor
def stop(self):
self.log.warning("stop input thread")
self.running = False
def run(self):
log.debug("starting input thread")
job_num = 0
while self.running and self.input_tube:
try:
job = self.beanstalk.reserve(self.input_tube, 3)
if job:
job_num += 1
body = job.body
resp = None
job.delete()
if self.processor is not None:
topic_id = None
try:
if type(self.processor).__name__ in (
'ExtractorProccessor',
'SingleSrcMergerProccessor'):
resp, topic_id = self.processor.do_task(body)
else:
resp = self.processor.do_task(body)
except Exception, e:
log.error("Process failed. " +
traceback.format_exc())
if resp is not None:
self.output_msg(resp, topic_id)
else:
self.log.debug(current_process().name +
" : no msg from : %s" % (self.input_tube))
except SocketError as e:
time.sleep(30)
self.log.error('beanstalk\tconnect\tfail\tstart\treconnect')
try:
self.beanstalk.reconnect()
self.out_beanstalk.reconnect()
self.log.error('beanstalk\treconnect\tsuccess')
except Exception as e:
self.log.error('beanstalk\treconnect\tfail')
#!/usr/bin/Python
# coding=utf-8
import sys
from thrift.protocol.TBinaryProtocol import TBinaryProtocol
from thrift.transport.TTransport import TMemoryBuffer
sys.path.append('..')
from bdp.i_crawler.i_extractor.ttypes import PageParseInfo
from i_util.pybeanstalk import PyBeanstalk
if __name__ == '__main__':
pybeanstalk = PyBeanstalk('101.201.100.58')
try:
rsp_info = PageParseInfo()
job = pybeanstalk.reserve('extract_info_ws')
# while True:
if job:
tMemory_o = TMemoryBuffer(job.body)
tBinaryProtocol_o = TBinaryProtocol(tMemory_o)
rsp_info.read(tBinaryProtocol_o)
d = vars(rsp_info)
print d
for k, v in d.items():
print k, v
job.delete()
except EOFError, e:
print e
from thrift.protocol.TBinaryProtocol import TBinaryProtocol
from thrift.transport.TTransport import TMemoryBuffer
sys.path.append('..')
from bdp.i_crawler.i_crawler_merge.ttypes import LinkAttr
from bdp.i_crawler.i_downloader.ttypes import DownLoadReq
from bdp.i_crawler.i_extractor.ttypes import PageParseInfo
from i_util.pybeanstalk import PyBeanstalk
if __name__ == '__main__':
pybeanstalk = PyBeanstalk('10.25.114.50')
try:
#link_info = DownLoadReq();
info = PageParseInfo()
while True:
#print pybeanstalk.stats_tube('download_req')
job = pybeanstalk.reserve('online_extract_info')
"""
tMemory_o = TMemoryBuffer(job.body)
tBinaryProtocol_o = TBinaryProtocol(tMemory_o)
info.read(tBinaryProtocol_o)
d = vars(info)
for k,v in d.items():
print k,v
"""
job.delete()
#break;
except EOFError, e:
print e
class InputThread(threading.Thread):
def __init__(self, conf, processor, proc_name= None):
threading.Thread.__init__(self)
self.daemon = True
self.running = True
self.beanstalk = PyBeanstalk(conf['beanstalk_conf']['host'], conf['beanstalk_conf']['port'])
self.out_beanstalk = PyBeanstalk(conf['beanstalk_conf']['host'], conf['beanstalk_conf']['port'])
self.input_tube = conf['beanstalk_conf']['input_tube']
self.output_tube = conf['beanstalk_conf']['output_tube']
self.log = conf['log']
if not self.log:
self.log = LogHandler("i_input_thread")
self.processor = processor
if self.processor is None:
self.log.error("Processor not given !")
raise Exception("Processor not given !")
self.processor_pool = ThreadPool(conf['server'].get("process_thread_num", 1),\
{},\
int(conf['server'].get("process_thread_num", 1))
)
self.wlock = threading.Lock()
def stop(self):
self.log.warning("stop input_thread")
self.running = False
self.processor_pool.join_all()
def run(self):
job_num = 0
self.running = True
while self.running and self.input_tube:
try:
job = self.beanstalk.reserve(self.input_tube, 3)
if not job is None:
job_num += 1
body = job.body
job.delete()
self.processor_pool.queue_task(self._on_task_start, (body,), self._on_task_finished)
except SocketError as e:
time.sleep(30)
self.log.error('beanstalk\tconnect\tfail\tstart\treconnect')
try:
self.beanstalk.reconnect()
self.out_beanstalk.reconnect()
self.log.error('beanstalk\treconnect\tsuccess')
except Exception as e:
self.log.error('beanstalk\treconnect\tfail')
except:
self.log.error("not msg from:%s\tresult:%s" % (self.input_tube, str(traceback.format_exc())))
def _on_task_start(self, task, **thread_locals):
result = None
try:
result = self.processor.do_task(task)
except Exception as e:
self.log.error(e.message)
return result
def _on_task_finished(self, (task), **thread_locals):
self.wlock.acquire()
if task and isinstance(task, basestring):
self._output_msg(task)
elif isinstance(task, list):
for message in task:
self._output_msg(message)
self.wlock.release()
#!/usr/bin/Python
# coding=utf-8
import sys
from thrift.protocol.TBinaryProtocol import TBinaryProtocol
from thrift.transport.TTransport import TMemoryBuffer
sys.path.append('..')
from bdp.i_crawler.i_downloader.ttypes import DownLoadRsp
from bdp.i_crawler.i_extractor.ttypes import PageParseInfo
from i_util.pybeanstalk import PyBeanstalk
if __name__ == '__main__':
pybeanstalk = PyBeanstalk('101.201.102.37')
try:
extractor_info = PageParseInfo()
body = pybeanstalk.reserve('extract_info').body
tMemory_o = TMemoryBuffer(body)
tBinaryProtocol_o = TBinaryProtocol(tMemory_o)
extractor_info.read(tBinaryProtocol_o)
print extractor_info
except EOFError, e:
print e
class InputThread(threading.Thread):
def __init__(self, beanstalk_conf, log=None, process_pool=None):
threading.Thread.__init__(self)
self.daemon = True
self.running = True
assert beanstalk_conf is not None
assert log is not None
assert process_pool is not None
self.beanstalk = PyBeanstalk(beanstalk_conf['host'], beanstalk_conf['port'])
self.out_beanstalk = PyBeanstalk(beanstalk_conf['host'], beanstalk_conf['port'])
self.input_tube = beanstalk_conf['input_tube']
self.output_tube = beanstalk_conf['output_tube']
self.log = log
if not self.log:
self.log = LogHandler("i_input_thread")
self.process_pool = process_pool
self.t_lock = threading.Lock()
def stop(self):
self.log.warning("stop input_thread")
self.running = False
try:
while True:
if self.process_pool.get_task_num() <= 0:
# if 'processor' in self.process_pool.thread_local_constructors:
# processor = self.process_pool.thread_local_constructors['processor'][1][1]
# self.log.warning("prepare call scheduler_processor to stop scheduler")
# processor.save_status()
break
else:
self.log.info("wait tasks be consumed over, wait 5s")
time.sleep(5)
self.beanstalk.__del__() # 关闭连接不再接受数据
except Exception as e:
self.log.error("stop input_thread fail")
self.log.exception(e)
def run(self):
job_num = 0
while self.running and self.input_tube:
try:
job = self.beanstalk.reserve(self.input_tube, 30)
if job is not None:
job_num += 1
body = job.body
job.delete()
self.process_pool.queue_task(self.__on_task_start, (body,), self.__on_task_finished)
task_num = self.process_pool.get_task_num()
if task_num >= 50:
self.log.info("place_processor\ttasks:%d" % task_num)
time.sleep(2)
else:
self.log.info("not msg from:%s" % self.input_tube)
except SocketError as e:
time.sleep(30)
self.log.error('beanstalk\tconnect\tfail\tstart\treconnect')
self.log.exception(e)
try:
self.beanstalk.reconnect()
self.out_beanstalk.reconnect()
self.log.error('beanstalk\treconnect\tsuccess')
except Exception as e:
self.log.error('beanstalk\treconnect\tfail')
self.log.exception(e)
except Exception as e:
self.log.error("not msg from:%s\tresult:" % self.input_tube)
self.log.exception(e)
@staticmethod
def __on_task_start(task, **thread_locals):
result = None
if 'profiler' in thread_locals:
thread_locals['profiler'].begin()
if 'processor' in thread_locals:
result = thread_locals['processor'].do_task(task)
return result
def __on_task_finished(self, (result), **thread_locals):
self.t_lock.acquire()
proccesor = None
if 'processor' in thread_locals:
proccesor = thread_locals['processor']
if 'profiler' in thread_locals:
thread_locals['profiler'].end()
if result and isinstance(result, basestring):
self.__output_msg(result, proccesor)
elif isinstance(result, list):
for message in result:
self.__output_msg(message, proccesor)
self.t_lock.release()
