0.5.7 waiting connections

[Igor Sysoev]

On Sat, 20 Jan 2007, [UTF-8] Jan ?lusarczyk wrote:

>>>> Is anyting in error_log at alert level ?
>> 
>> In general the connections should not grow in time.
>
> Hi Igor
> Thank you for your help. After the night I have the same problem - number of 
> connections going steadily up. But this time I have more data:
> 1. The control process died yesterday evening. I have the following data in 
> my warn.log:
>
> 2007/01/19 22:09:22 [alert] 28060#0: epoll_wait() failed (22: Invalid 
> argument)
>
> Then 860.000 lines of almost the same (10 seconds at approx 60.000 per 
> second). Then:
>
> 2007/01/19 23:27:02 [alert] 6755#0: epoll_wait() failed (22: Invalid 
> argument)

It's strange.

> for a few hundred lines. And finally:
>
> 2007/01/19 23:27:03 [alert] 27079#0: worker process 6755 exited on signal 11
> 2007/01/19 23:30:47 [alert] 976368160#0: open socket #15 left in 23 
> connection, aborting
> 2007/01/19 23:30:47 [alert] 27079#0: close() channel failed (9: Bad file 
> descriptor)
> 2007/01/19 23:30:47 [alert] 27079#0: close() channel failed (9: Bad file 
> descriptor)
> 2007/01/19 23:30:47 [alert] 27079#0: sendmsg() failed (9: Bad file

> 2007/01/19 23:16:23 [alert] 5516#0: *114759 zero size buf in output t:0 r:0 
> f:1 00000000 00000000-00000000 081DEB30 0-0 while sending response to client, 
> client: 81.210.63.229, server: www.babyboom.pl, URL: 
> "/fileadmin/_temp_/Memo_tabl_do50.pdf", host: "www.babyboom.pl"

> Any idea what may be the reason? Any way I can help to further debug this 
> issue?

What Linux verison do you use ?
What does "nginx -V" show ?


Igor Sysoev
http://sysoev.ru/en/