Bug 23300 - Asynchronous AcceptEx failed
Summary: Asynchronous AcceptEx failed
Status: CLOSED DUPLICATE of bug 8325
Alias: None
Product: Apache httpd-2
Classification: Unclassified
Component: mpm_winnt (show other bugs)
Version: 2.0.47
Hardware: PC All
: P3 critical (vote)
Target Milestone: ---
Assignee: Apache HTTPD Bugs Mailing List
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2003-09-21 08:17 UTC by Yongxiang Peng
Modified: 2005-03-20 17:06 UTC (History)
1 user (show)



Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Yongxiang Peng 2003-09-21 08:17:35 UTC
logs... after start service... ->>>

[Sun Sep 21 15:05:45 2003] [notice] Parent: Created child process 1232
[Sun Sep 21 15:05:45 2003] [debug] mpm_winnt.c(505): Parent: Sent the 
scoreboard to the child
[Sun Sep 21 15:05:45 2003] [notice] Child 1232: Child process is running
[Sun Sep 21 15:05:45 2003] [info] Parent: Duplicating socket 284 and sending it 
to child process 1232
[Sun Sep 21 15:05:45 2003] [debug] mpm_winnt.c(426): Child 1232: Retrieved our 
scoreboard from the parent.
[Sun Sep 21 15:05:45 2003] [debug] mpm_winnt.c(623): Parent: Sent 1 listeners 
to child 1232
[Sun Sep 21 15:05:45 2003] [debug] mpm_winnt.c(582): Child 1232: retrieved 1 
listeners from parent
[Sun Sep 21 15:05:45 2003] [notice] Child 1232: Acquired the start mutex.
[Sun Sep 21 15:05:45 2003] [notice] Child 1232: Starting 2 worker threads.
[Sun Sep 21 15:05:45 2003] [debug] child.c(695): Child 1232: Worker thread 0 
starting.
[Sun Sep 21 15:05:45 2003] [debug] child.c(695): Child 1232: Worker thread 1 
starting.
[Sun Sep 21 15:05:45 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
[Sun Sep 21 15:07:03 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
[Sun Sep 21 15:07:03 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
[Sun Sep 21 15:07:03 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
[Sun Sep 21 15:07:03 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
[Sun Sep 21 15:07:03 2003] [debug] child.c(547): (OS 10038)take a operation on 
a no socket.  : winnt_accept: AcceptEx failed due to early client disconnect. 
Reallocate the accept socket and try again.
Comment 1 André Malo 2003-09-21 19:34:46 UTC
Do you have a (personal?) firewall or virusscanner running?
Comment 2 Yongxiang Peng 2003-09-23 05:00:06 UTC
Neithor firewall nor virusscanner is running!
And not take any connecting operation.
Comment 3 David Ramalho 2003-09-24 11:42:22 UTC
  I leapt from version 2.0.36 to 2.0.47 and the problem started to show up, in 
a massive way. On debug level we get a few other lines, but their probably 
related. Also derived from that, I guess, was the problem of the server 
starting to get REAAAAAAAAAAAAAALY slow, with some many 'exceptions' coming 
from the OS, so my solution was to disable the KeepAlive, witch, I we think 
about it, is what's causing the problem, in the sense that he needs to keep the 
connection, during witch time, some Network stuff happens and the OS (or 
Apache) can't keep up. 
  Like I said, turning KeepAlive off, totally cleared my error log of that sort 
of warning, BUT ... and this is probably a different bug, my Apache keeps 
getting killed ... some times it seems to start by himself, some other times de 
service dies and Windows picks it up again in one minute (I changed the configs 
of the service). There is no apparent reason for this, the only module I'm 
running is PHP 4.3.3 and, like I said, the OS warning about Network problems 
disappeared, so I'm baffled. The previous version I had (2.0.36) ran for months 
with the same amount of requests I have now.
Comment 4 David Ramalho 2003-09-24 11:48:14 UTC
From time to time (a few hours) this happens on my server : 

[Wed Sep 24 12:05:43 2003] [notice] Parent: child process exited with status 
4294967295 -- Restarting.
[Wed Sep 24 12:05:43 2003] [notice] Parent: Created child process 1352
[Wed Sep 24 12:05:44 2003] [notice] Child 1352: Child process is running
[Wed Sep 24 12:05:44 2003] [notice] Child 1352: Acquired the start mutex.
[Wed Sep 24 12:05:44 2003] [notice] Child 1352: Starting 15 worker threads.

In this case he got up by himself (aparently)

The rest of the times he does this : 
[Wed Sep 24 12:22:49 2003] [notice] Parent: child process exited with status 
4294967295 -- Restarting.
[Wed Sep 24 12:23:50 2003] [notice] Parent: Created child process 2016
[Wed Sep 24 12:23:50 2003] [notice] Child 2016: Child process is running
[Wed Sep 24 12:23:50 2003] [notice] Child 2016: Acquired the start mutex.
[Wed Sep 24 12:23:50 2003] [notice] Child 2016: Starting 15 worker threads.

Same thing, but apparently this time the service stops, and 60 seconds latter, 
mr. Win2k starts it.

Don't know why ... that's the thing, no help in the error log, even in debug 
mode .
Comment 5 Jeff Trawick 2003-10-10 12:39:54 UTC
fix the component
fix the OS (seems to be Win2K)
Comment 6 Cliff Woolley 2003-10-11 01:14:29 UTC
Okay, so just to reiterate... you do NOT have McAfee installed, right?  It 
doesn't have to be running; if it's just _installed_, it will cause this 
problem.  See Bug 8235. 
Comment 7 David Ramalho 2003-10-11 12:15:50 UTC
  You did good in reiterating, I think I have McAfee Viruscan installed. I 
forgot to mention because I was thinking only in Firewalls, so this would cause 
the socket fail? From a long view seems to be rather ... well ... unrelated, 
but I trust experience here :) . Never the less, is it something _global_ to 
win2k (Aparently XP too), or will it be addressed by Apache in some way?
  As for the crashing part, there's another bug report
http://nagoya.apache.org/bugzilla/show_bug.cgi?id=18963
  that shows the same behavior on the child process beeing restarted. 
  
  As for this particular bug, as soon as I can I'll try ... not deactivate, but 
unninstall McAfee to see if it helps ... although, it's not a solution I 
particulary like :)

  Thanks
Comment 8 Cliff Woolley 2003-10-11 21:03:46 UTC
My crossref should have said Bug 8325, not 8235.  oops.  :)
Comment 9 Cliff Woolley 2003-10-11 21:05:09 UTC

*** This bug has been marked as a duplicate of 8325 ***