Django服务器100％cpu没有明显的原因

Question

这是一个运行在Linux（CentOS我认为）使用Python 2.6.6和Django 1.2.3的服务器。

会发生什么事是运行django的python进程突然开始使用100％cpu，直到它重新启动。这只是最近才发生过两次，而且在不到一个月前才开始这样做。我在大约7个月内没有对代码做过任何大的修改。

查看控制台的输出，它没有任何极端的用法。在10分钟内只有大约10个查询导致我相信它开始使用100％cpu。打印出来的唯一错误是一个破损的管道错误，我想可能是某人在缓慢关闭连接后使用它。

我重新运行了所有查询，这些查询都是在它放缓的时候出现的，并且它们都工作得很好，没有任何问题。

服务器本身在某种意义上仍然有效，但速度非常慢。我有一系列的测试，我每天运行他们通常需要7分钟左右，但是当它的速度变慢时，可能需要2-3小时。

如果有人有任何想法，我将非常感激。

另外，正如你可能会注意到，如果有人可以推荐一个关于如何监控这些活动的良好实践，那么当谈到这些问题时，我是一个相当新手。

谢谢你的时间！

下面是我所提到的输出，它开始100％的CPU时间是下午4点〜

[23/Jul/2012 15:49:55] "GET /CFXsearch/?n=&v=all&e=2012Week30&c=all&r=all&p=all&run=all&pl=all&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 67228 
[23/Jul/2012 15:50:00] "GET /CFXsearch/?n=&v=all&e=2012Week30&c=all&r=all&p=all&run=all&pl=RH5&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 33346 
[23/Jul/2012 15:50:05] "GET /CFXsearch/?n=&v=all&e=2012Week30&c=all&r=all&p=all&run=all&pl=SLES10&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 33394 
[23/Jul/2012 15:54:48] "GET /CFXsearch/?n=&v=all&e=2012Week30&c=all&r=all&p=all&run=all&pl=SLES11&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 33394 
[23/Jul/2012 15:54:53] "GET /results/?n=&e=2012Week30&c=TGTest&pl=SLES11&p=single&p=defined&p=double&run=default&run=hpmpi&run=mpich&run=mpich2&run=Platform&run=pvm%20parallel&run=serial&sort=sortResult&d=&y=&submitOption=latestsearch&Soft=CFX HTTP/1.1" 200 19350 
[23/Jul/2012 15:54:57] "GET /results/?n=&e=2012Week30&c=turboexamples&pl=SLES11&p=single&p=defined&p=double&run=default&run=hpmpi&run=mpich&run=mpich2&run=Platform&run=pvm%20parallel&run=serial&sort=sortResult&d=&y=&submitOption=latestsearch&Soft=CFX HTTP/1.1" 200 36729 
[23/Jul/2012 15:59:40] "GET/HTTP/1.1" 200 11111 
[23/Jul/2012 15:59:40] "GET /site_media/style.css HTTP/1.1" 304 0 
[23/Jul/2012 15:59:45] "GET /CFXsearch/ HTTP/1.1" 200 25637 
[23/Jul/2012 15:59:45] "GET /site_media/jquery-1.2.6.min.js HTTP/1.1" 304 0 
[23/Jul/2012 15:59:45] "GET /site_media/sorttable.js HTTP/1.1" 304 0 
[23/Jul/2012 16:00:04] "GET /CFXsearch/?n=&v=14.5&e=all&c=solver54&r=all&p=all&run=all&pl=all&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 402737 
[23/Jul/2012 16:00:19] "GET /results/?n=&e=2012Week29&c=solver54&pl=SLES11&p=single&p=defined&p=double&run=default&run=hpmpi&run=mpich&run=mpich2&run=Platform&run=pvm%20parallel&run=serial&sort=sortResult&d=&y=&submitOption=latestsearch&Soft=CFX HTTP/1.1" 200 1557488 
[23/Jul/2012 16:02:48] "GET /CFXsearch/?n=&v=14.5&e=all&c=solver54&r=all&p=all&run=all&pl=all&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=ex&sort=sortName&Soft=CFX HTTP/1.1" 200 408388 
[23/Jul/2012 16:03:01] "GET /CFXsearch/?n=&v=14.5&e=all&c=solver54&r=all&p=all&run=all&pl=all&m=all&o=all&d1=&d2=&submitOption=comparing&compareBy=plat&sort=sortName&Soft=CFX HTTP/1.1" 200 402737 
Traceback (most recent call last): 
    File "/home/install2/testingDatabase/lib/python2.6/site-packages/django/core/servers/basehttp.py", line 281, in run 
    self.finish_response() 
    File "/home/install2/testingDatabase/lib/python2.6/site-packages/django/core/servers/basehttp.py", line 321, in finish_response 
    self.write(data) 
    File "/home/install2/testingDatabase/lib/python2.6/site-packages/django/core/servers/basehttp.py", line 400, in write 
    self.send_headers() 
    File "/home/install2/testingDatabase/lib/python2.6/site-packages/django/core/servers/basehttp.py", line 465, in send_headers 
    self._write(str(self.headers)) 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/socket.py", line 318, in write 
    self.flush() 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/socket.py", line 297, in flush 
    self._sock.sendall(buffer(data, write_offset, buffer_size)) 
error: [Errno 32] Broken pipe 
Traceback (most recent call last): 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/SocketServer.py", line 560, in process_request_thread 
    self.finish_request(request, client_address) 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/SocketServer.py", line 322, in finish_request 
    self.RequestHandlerClass(request, client_address, self) 
    File "/home/install2/testingDatabase/lib/python2.6/site-packages/django/core/servers/basehttp.py", line 562, in __init__ 
    BaseHTTPRequestHandler.__init__(self, *args, **kwargs) 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/SocketServer.py", line 618, in __init__ 
    self.finish() 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/SocketServer.py", line 661, in finish 
    self.wfile.flush() 
    File "/home/install2/testingDatabase/Python-2.6.6/Lib/socket.py", line 297, in flush 
    self._sock.sendall(buffer(data, write_offset, buffer_size)) 
error: [Errno 32] Broken pipe 
[23/Jul/2012 16:09:59] "GET /PolyflowSummary/ HTTP/1.1" 200 7561 
[23/Jul/2012 16:17:42] "GET /?soft=CFX HTTP/1.1" 200 11112 
[23/Jul/2012 16:17:44] "GET /?soft=CFX HTTP/1.1" 200 11112 
[23/Jul/2012 16:18:06] "GET /site_media/style.css HTTP/1.1" 200 432 
[23/Jul/2012 16:18:23] "GET /site_media/style.css HTTP/1.1" 200 432 
[23/Jul/2012 16:18:23] "GET /site_media/favicon.ico HTTP/1.1" 200 1718 

Answer 1

您可以连接到该进程并附加调试器。我之前做过这个，这非常有用。 my full notes are here，但删节版本是：

• 安装this使得GDB “理解” 蟒

• 使用gdb -p PID（PID从ps或类似）

• 连接产生在gdb栈跟踪你会发现你正在吃掉CPU。

原信用证 - Showing the stack trace from a running Python application（事实上，打字一切之后，也许这就是链接的问题愚弄的人我想问题是不同的，即使答案是一样的...？）