nginx-0.7.53/Windows select error and dead loop


#1

Hello,

I started 0.7.53 Windows build hours ago and it come into crazy dead
loop minutes later, these have been occured two times. The CPU is 100%
since I started two worker, and the error.log file has been increasing
to nearly 8GB with “select error” messages.

Here are my nginx.conf:

---- nginx.conf begin ----

worker_processes 2;
error_log D:/diogin/http/site/logs/error.log;
pid D:/diogin/http/site/temp/nginx.pid;
events {
worker_connections 1024;
}
http {
include D:/diogin/soft/dat/nginx/mime.types;
default_type application/octet-stream;
log_format main ‘$remote_addr [$time_local] “$request”
$status’;
access_log D:/diogin/http/site/logs/access.log main;
sendfile on;
keepalive_timeout 65;
server {
listen 83;
server_name site.example.net;
gzip on;
gzip_types text/css application/x-javascript;
location / {
root D:/example/site;
index index.html index.htm;
rewrite ^(.*)-[0-9]+.(css|js|gif|jpe?g|png|bmp|swf|ico)$
$1.$2 break;
add_header Cache-Control max-age=2952000;
}
}
}

---- nginx.conf end ----

and these are the errors from error.log:

---- error.log begin ----

2009/04/28 13:47:00 2448#1732: select event fd:120 ev:0 is already set
while accepting new connection on 0.0.0.0:83
2009/04/28 15:28:37 1136#2784: select event fd:120 ev:0 is already set
while accepting new connection on 0.0.0.0:83
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)
2009/04/28 15:28:58 1136#2784: select() failed (10022: An invalid
argument was supplied)

---- error.log end ----

Thanks.

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,1495,1495#msg-1495


#2

I checked netstat -an, and here are the results:

TCP 0.0.0.0:83 0.0.0.0:0 LISTENING
TCP 0.0.0.0:83 0.0.0.0:0 LISTENING

It seems the two worker processes both try to listen on the same socket,
which generates the error “select event fd:120 ev:0 is already set while
accepting new connection on 0.0.0.0:83”?

Am I missing anything or is my nginx.conf misconfigured?

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,1495,1496#msg-1496


#3

diogin wrote:

I checked netstat -an, and here are the results:

TCP 0.0.0.0:83 0.0.0.0:0 LISTENING
TCP 0.0.0.0:83 0.0.0.0:0 LISTENING

It seems the two worker processes both try to listen on the same socket,
which generates the error “select event fd:120 ev:0 is already set while
accepting new connection on 0.0.0.0:83”?

Am I missing anything or is my nginx.conf misconfigured?

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,1495,1496#msg-1496

oh, I have posted this bug before you
http://www.ruby-forum.com/topic/184732#808203
http://www.ruby-forum.com/topic/184732#808773

I think worker_processes set only 1 now


#4

Igor S. wrote:

On Tue, Apr 28, 2009 at 11:10:17AM +0200, Chris Wan wrote:

Am I missing anything or is my nginx.conf misconfigured?

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,1495,1496#msg-1496

oh, I have posted this bug before you
http://www.ruby-forum.com/topic/184732#808203
http://www.ruby-forum.com/topic/184732#808773

I think worker_processes set only 1 now

No, this is not workers problem, this is nginx’s bug.
Could anyone reproduce the bug with debug log ?

I have sent the letter to you.


#5

On Tue, Apr 28, 2009 at 11:10:17AM +0200, Chris Wan wrote:

Am I missing anything or is my nginx.conf misconfigured?

Posted at Nginx Forum:
http://forum.nginx.org/read.php?2,1495,1496#msg-1496

oh, I have posted this bug before you
http://www.ruby-forum.com/topic/184732#808203
http://www.ruby-forum.com/topic/184732#808773

I think worker_processes set only 1 now

No, this is not workers problem, this is nginx’s bug.
Could anyone reproduce the bug with debug log ?