def __init__(self, beanstalk_conf, log=None, process_pool=None):
threading.Thread.__init__(self)
self.daemon = True
self.running = True
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.wlock = threading.Lock()
class InputThread(threading.Thread):
def __init__(self, beanstalk_conf, log=None, process_pool=None):
threading.Thread.__init__(self)
self.daemon = True
self.running = True
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.wlock = threading.Lock()
def stop(self):
self.log.warning("stop input_thread")
self.running = False
proccesor = None
try:
while True:
if self.process_pool.get_task_num() == 0:
if self.process_pool.thread_local_constructors.has_key('processor'):
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, e:
self.log.error("stop input_thread fail:%s" % e.message)
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 init_log(self, conf, level=logging.DEBUG, console_out=False, name=None):
if isinstance(level, basestring):
level = level.lower()
if level == "debug":
level = logging.DEBUG
elif level == "info":
level = logging.INFO
elif level == "warning":
level = logging.WARNING
elif level == "error":
level = logging.ERROR
else:
level = logging.DEBUG
if name is None:
name = conf['server']['name']
self.log = LogHandler(name, level, console_out=console_out)
selector = CrawlSelector(conf['log'], conf['selector_conf'],
conf['beanstalk_conf'], None)
selector.start()
while True:
time.sleep(20)
selector.stop()
if __name__ == '__main__':
try:
file_path = './scheduler.toml'
opt, args = getopt.getopt(sys.argv[1:], 'f:', ['help'])
for name, value in opt:
if name == "-f":
file_path = value
elif name in ("-h", "--help"):
usage()
sys.exit()
else:
assert False, "unhandled option"
with open(file_path, 'rb') as config:
conf = pytoml.load(config)
conf['log'] = LogHandler(conf['server']['name'] +
str(conf['server']['port']))
main(conf)
except getopt.GetoptError:
sys.exit()
except Exception, e:
self.log.error("report activate failed, because %s" % e.message)
time.sleep(self.EXPIREDS)
def stop(self):
self.running = False
if __name__ == "__main__":
import logging
from i_util.logs import LogHandler
# log = LogHandler(config.logname, logging.DEBUG)
log = LogHandler('test', logging.DEBUG)
conf = {
"backend": {"host": "101.201.102.37",
"port": 6379,
"password": "******"},
"server_name": "test",
"server": {"host": "127.0.0.1",
"port": 1001},
"log": log
}
obj = HeartbeatThread("extractor", conf=conf)
obj.run()
'port': 6379,
'database': 8,
'proxy_name': 'proxies',
'proxy_test_available': 24 #hours
}
SERVER = {'port': 8571, 'host': '0.0.0.0', 'debug': False}
MONGODB = {
'host': '172.17.1.119',
'port': 40042,
'database': 'task_collect',
'username': '******',
'password': '******'
}
final_data = {
'host': '172.17.1.119',
'port': 40042,
'name': 'app_data',
'username': '******',
'password': '******'
}
realtime_crawl = {
'api': 'http://172.18.180.225:9823/gsxt_info',
'query': 'company'
}
from i_util.logs import LogHandler
log = LogHandler("manager", loglevel=logging.INFO)
def main():
logger = LogHandler('select_webpage')
reload(sys)
sys.setdefaultencoding("utf-8")
import os
import json
import traceback
import time
from thrift.protocol.TBinaryProtocol import TBinaryProtocol
from thrift.transport.TTransport import TMemoryBuffer
sys.path.append('..')
from i_util.pybeanstalk import PyBeanstalk
from bdp.i_crawler.i_downloader.ttypes import DownLoadReq
from i_util.logs import LogHandler
import config
log = LogHandler('re_crawler', console_out=True)
beanstalk_conf = config.beanstalk_conf
beanstalk_client = PyBeanstalk(beanstalk_conf['host'], beanstalk_conf['port'])
def req_to_string(req):
str_req = ""
try:
tMemory_b = TMemoryBuffer()
tBinaryProtocol_b = TBinaryProtocol(tMemory_b)
req.write(tBinaryProtocol_b)
str_req = tMemory_b.getvalue()
except:
log.error('crawled_failt\terror:%s' % (traceback.format_exc()))
return str_req
log.warning("cann't write DownLoadRsp to string")
return str_page_info
def put_beanstalked(beanstalk_conf, log, rsp):
beanstalk = PyBeanstalk(beanstalk_conf['host'], beanstalk_conf['port'])
tube = beanstalk_conf['input_tube']
str_page_info = to_string(log, rsp)
try:
beanstalk.put(tube, str_page_info)
log.info('beanstalk\turl:%s\ttube:%s' % (rsp.url, tube))
except Exception as e:
log.info('beanstalk put error url:' + rsp.url + '\ttube:' + tube)
if __name__ == "__main__":
url = 'dsfsdf'
redirect_url = 'sdfdfsfsfs'
status = 1
http_code = 200
rsp = DownLoadRsp(url=url,
redirect_url=redirect_url,
status=status,
http_code=http_code)
beanstalk_conf = {}
beanstalk_conf['host'] = '101.201.102.37'
beanstalk_conf['port'] = 11300
beanstalk_conf['input_tube'] = 'download_rsp_test'
log = LogHandler('download_test')
put_beanstalked(beanstalk_conf, log, rsp=rsp)
# -*- coding: utf-8 -*-
from conf import config
# logging是线程安全的
import logging
# logging.basicConfig(level = logging.DEBUG)
from i_util.logs import LogHandler
log = LogHandler(config.server['name']+str(config.server['port']), logging.INFO)
import MySQLdb
def get_mysqldb():
# 1 mysql init
mysql_db = MySQLdb.connect(
host=config.MYSQL['host'],
port=config.MYSQL['port'],
user=config.MYSQL['username'],
passwd=config.MYSQL['password'],
db=config.MYSQL['database'],
charset='utf8'
)
return mysql_db
def dbrecord_to_dict(descriptions, record):
'''mysql元组记录转换为字典'''
return {
desc[0]: field_value
for desc, field_value in zip(descriptions, record)
}
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()
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()
host = '127.0.0.1'
port = 12400
server_thread_num = 1
server_process_num = 1
process_thread_num = 1
# beanstalk_conf = {
# 'host': '101.201.102.37', # 线上beanstalk内网IP
# 'port': 11300, # 线上beanstalk内网port
# 'input_tube': 'extract_to_crawl',
# 'output_tube': 'scheduler_info',
# }
beanstalk_conf = {
'host': '127.0.0.1', # 线上beanstalk内网IP
'port': 11300, # 线上beanstalk内网port
'input_tube': 'extract_to_crawl',
'output_tube': 'scheduler_info',
}
mysql_host = '101.201.102.37'
mysql_user = '******'
mysql_passwd = 'haizhi@)'
mysql_db = 'cmb_crawl'
download_output_tube = 'czj_download_rsp'
from i_util.logs import LogHandler
log = LogHandler("crawler_merge")
# -*- coding:utf-8 -*-
import threading
import traceback
import MySQLdb
from i_util.logs import LogHandler
log = LogHandler('redisdb')
class PyMySQL(object):
FETCH_ONE = 0
FETCH_MANY = 1
FETCH_ALL = 2
SELECT_SQL_FORMAT = "SELECT %s FROM `%s` WHERE %s"
INSERT_SQL_FORMAT = "INSERT INTO `%s` (%s) VALUES (%s)"
UPDATE_SQL_FORMAT = "UPDATE `%s` SET %s WHERE %s"
DETELE_SQL_FORMAT = "DELETE FROM `%s` WHERE %s"
def __init__(self, host, post, database, username, password):
self.conn = None
self.host = host
self.post = post
self.database = database
self.username = username
self.password = password
self._connect()
# signal.signal(signal.SIGINT, signal_process)
# signal.signal(signal.SIGQUIT, signal_process)
# signal.signal(signal.SIGUSR1, lambda a, b: profiling_signal_handler("downloader", a, b))
file_path = 'downloader_smart_test.toml'
opt, args = getopt.getopt(sys.argv[1:], 'f:', ['help'])
for name, value in opt:
if name == "-f":
file_path = value
elif name in ("-h", "--help"):
usaget()
sys.exit()
else:
assert False, "unhandled option"
if file_path is None or file_path.strip() == '':
raise Exception('配置文件路径错误..')
with open(file_path, 'rb') as fp:
config = pytoml.load(fp)
logger_name = config["local_server"].get('name') + str(
config["local_server"].get('port'))
logger = LogHandler(logger_name)
config['log'] = logger
logger.info(logger_name)
logger.info('开始启动服务....')
main(config)
except getopt.GetoptError:
raise Exception(
'参数读取错误...argv = {argv}'.format(argv=" ".join(sys.argv)))
