def run(self):
try:
self.spiderManager.Clock_Sta = ThreadSta.Work
self.check_date()
except Exception:
report_logger.error("Clock_Thread 线程出错")
self.spiderManager.Clock_Sta = ThreadSta.Error
def run(self):
self.spiderManager.Monitor_Sta = ThreadSta.Work
try:
self._run()
except Exception:
self.spiderManager.Monitor_Sta = ThreadSta.Error
report_logger.error("MonitorThread 出错")
def run(self):
self.spiderManager.Parse_Sta = ThreadSta.Work
try:
self._run()
except Exception as exc:
report_logger.error("ParseThread:\n" + traceback.format_exc())
self.spiderManager.Parse_Sta = ThreadSta.Error
def run(self):
self.spiderManager.Fetch_Sta = ThreadSta.Work
try:
self._run()
except Exception as exc:
report_logger.error("FetchThread ", exc)
self.spiderManager.Fetch_Sta = ThreadSta.Error
def run(self):
self.spiderManager.TaskAllot_Sta = ThreadSta.Work
try:
self._run()
except Exception:
report_logger.error("TaskAllotThread 出错")
self.spiderManager.fetch_queue.is_valid = False
self.spiderManager.TaskAllot_Sta = ThreadSta.Error
def check_thread_error(self):
for sta in [
self.Fetch_Sta, self.Parse_Sta, self.Save_Sta, self.Clock_Sta,
self.TaskAllot_Sta, self.Monitor_Sta
]:
if sta == ThreadSta.Error:
report_logger.error("运行报错,且退出")
return True
return False
def start_work_and_wait_done(self,
fetcher_num=10,
parser_num=1,
monitor_time=5):
"""
开启爬取
:param monitor_time: 监视线程报告间隔
:param fetcher_num: 爬取线程数
:param parser_num: 解析进程数
"""
# 时钟线程
clock_thread = ClockThread(self.schedule, self.spiderManager)
# 监视线程
monitor_thread = MonitorThread(self.client,
self.spiderManager,
sleeptime=monitor_time)
# 任务调度线程
task_allot_thread = TaskAllot(self.client, self.schedule,
self.spiderManager)
# 爬取线程
if self.proxy:
print("启用 代理模式")
fetch_thread = FetchThreadProxy(self.client, self.fetcher,
fetcher_num, self.spiderManager)
else:
fetch_thread = FetchThread(self.client, self.fetcher, fetcher_num,
self.spiderManager)
# 解析线程
parse_thread = ParseThread(self.client, self.parser, parser_num,
self.spiderManager)
# 存储线程
save_thread = SaveThread(self.saver, self.spiderManager)
for thread in [
clock_thread, task_allot_thread, fetch_thread, parse_thread,
save_thread, monitor_thread
]:
thread.start()
while True:
if self.is_all_task_done(
) or self.spiderManager.check_thread_error():
self.spiderManager.finish_all_threads()
break
time.sleep(1)
for thread in [
fetch_thread, parse_thread, save_thread, monitor_thread
]:
thread.join()
report_logger.error("任务全部完成")
def _run(self):
while True:
time.sleep(self._sleeptime)
self._init_time += self._sleeptime
m, s = divmod(self._init_time, 60)
h, m = divmod(m, 60)
fetch_rec, fetch_fin = self.spiderManager.fetch_queue.get_count()
parse_rec, parse_fin = self.spiderManager.parse_queue.get_count()
save_rec, save_fin = self.spiderManager.save_queue.get_count()
dif_fetch = fetch_fin - self._last_fetch_num
dif_parse = parse_fin - self._last_parse_num
dif_save = save_fin - self._last_save_num
info = "idle_task_num=%d " % self.client.get_idle_tasks_size()
info += "fetch=(%d/%d,%d) " % (fetch_fin, fetch_rec, dif_fetch)
info += "parse=(%d/%d,%d) " % (parse_fin, parse_rec, dif_parse)
info += "save=(%d/%d,%d) " % (save_fin, save_rec, dif_save)
info += "times=%02d:%02d:%02d" % (h, m, s)
self._last_fetch_num = fetch_fin
self._last_parse_num = parse_fin
self._last_save_num = save_fin
montion_logger.debug(info)
if self.DEBUG is False and dif_fetch == 0 and dif_parse == 0 and dif_save == 0:
self.ErrorCount += 1
if self.ErrorCount >= self._IdleTime:
self.ErrorCount = 0
self.spiderManager.Monitor_Sta = ThreadSta.Error
report_logger.error("MonitorThread 长时间无进展")
break
else:
self.ErrorCount = 0
if self.spiderManager.Monitor_Sta == ThreadSta.Finish:
break