Active waiting?

Alvaro Gamez alvaro.gamez at hazent.com
Mon Mar 23 17:06:26 GMT 2015


Hi again, I got some interesting new information

2015-03-23 17:20 GMT+01:00 Alvaro Gamez <alvaro.gamez at hazent.com>:

>  --- modulename: threading, funcname: wait
>
> threading.py(235):         print "XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX ", timeout
> XXXXXXXXXXXXXXXXXXXXXXXXXXXXXX  None
> threading.py(236):         if not self._is_owned():
>  --- modulename: threading, funcname: _is_owned
> threading.py(228):         if self.__lock.acquire(0):
> threading.py(232):             return True
> threading.py(238):         waiter = _allocate_lock()
> threading.py(239):         waiter.acquire()
> threading.py(240):         self.__waiters.append(waiter)
> threading.py(241):         saved_state = self._release_save()
>  --- modulename: threading, funcname: _release_save
> threading.py(220):         self.__lock.release()           # No state to
> save
> threading.py(242):         try:    # restore state no matter what (e.g.,
> KeyboardInterrupt)
> threading.py(243):             if timeout is None:
> threading.py(244):                 waiter.acquire()
> threading.py(255):                 while True:
> threading.py(256):                     gotit = waiter.acquire(0)
> threading.py(257):                     if gotit:
> threading.py(259):                     remaining = endtime - _time()
> threading.py(260):                     if remaining <= 0:
> threading.py(262):                     delay = min(delay * 2, remaining,
> .05)
> threading.py(263):                     _sleep(delay)
>
>
After placing some more XXX and the like lines (as I told I'm not
experiencied with python, I'm sure there's a better way to do this), it
seems that I've triggered some optimizations out that provide me a more
complete trace.

threading.py(258):                     print "ZZZZZZZZZZZZZZZZZZZZZZZZZ ",
timeout
ZZZZZZZZZZZZZZZZZZZZZZZZZ  1740

That threading invocation with a timeout of 1740 matches the value of
29minutes*60seconds of keepalive parameter.

I've applied attached patch to allow a keepalive time of 0 seconds (which
internally equates to an infinite wait) and now CPU consumption is 0 and no
wakeups are notified by powertop, so I think this is definitely the thing
that is making offlineimap consume CPU while IDLING.

Of course attached patch is not a solution, because that 29 minutes is a
recommendation of the RFC for IDLE connections, so this is a timeout that
should definitely exist, and the one to blame here is Event.wait().

Best regards

diff --git a/offlineimap/repository/IMAP.py b/offlineimap/repository/IMAP.py
index 68c8e33..c2c279b 100644
--- a/offlineimap/repository/IMAP.py
+++ b/offlineimap/repository/IMAP.py
@@ -72,7 +72,7 @@ class IMAPRepository(BaseRepository):
         return self.getconfboolean("holdconnectionopen", 0)

     def getkeepalive(self):
-        num = self.getconfint("keepalive", 0)
+        return self.getconfint("keepalive", 0)
         if num == 0 and self.getidlefolders():
             return 29*60
         else:

-- 
Álvaro Gámez Machado
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://alioth-lists.debian.net/pipermail/offlineimap-project/attachments/20150323/ba866632/attachment-0003.html>


More information about the OfflineIMAP-project mailing list