urllib2不是线程安全的，并且opener更是共享的全局变量，如下urlib2.py(python2.7.5)代码：

#/usr/lib64/python2.7/urllib2.py
122 _opener = None                                                                                                                                   
123 def urlopen(url, data=None, timeout=socket._GLOBAL_DEFAULT_TIMEOUT):                                                                             
124     global _opener                                                                                                                               
125     if _opener is None:                                                                                                                          
126         _opener = build_opener()                                                                                                                 
127     return _opener.open(url, data, timeout)                                                                                                      
128                                                                                                                                                  
129 def install_opener(opener):                                                                                                                      
130     global _opener                                                                                                                               
131     _opener = opener

如果要在多线程里使用urllib需要保证install_opener不会被在多线程中同时调用，最好是在主线程安装opener。

如果要在多个线程中使用不同的opener，除了多进程外，还可以在各个子线程内创建自己的opener，然后使用各自opener的open函数进行url访问。如下：

# 不同的opener作为线程处理函数的参数，并通过urllib2.build_opener创建线程自己的opener。
def thread_func(opener):
    thread_opener = urllib2.build_opener(opener, other_default_opener)
    ret = thread_opener.open('http://www.baidu.com')
    # ...

最后针对这类url访问的任务比较推荐使用urlib3和requests，一方面线程安全，另一方面使用简单。