Strange 'too many open files' errors with nginx 0.6.32

hi,i’m seeing a strange ‘too many open files’ error with nginx (which
I’m
sure many of you have seen before):

2008/11/17 02:15:02 [alert] 21748#0: accept() failed (24: Too many open
files) while accepting new connection on 67.228.78.114:80

strangely enough, I have the following line in /etc/security/limits:

  • hard nofile 524288

and ulimit -n returns the same number. I have 2 workers set to a max
of 262144 connections. upon receiving that error (and after
past occurrences) I’ve run lsof and gotten < 10k sockets. additionally,
I
have a nearly identically configured machine (running twisted) where
each
process is handling ~15k connections. so I’m a loss as to what’s going
on
here (and suspect some subtle bug in nginx or in my server
configuration.
any help would be greatly appreciated!

oh, and one more thing, restarting nginx seems to fix everything (until
the
error pops up again a day or so later :slight_smile:

Hi there,

Add

worker_rlimit_nofile 10240;

to your nginx configuration should fix this problem

Regards,

Glen L.

From: [email protected] [mailto:[email protected]] On Behalf Of
Arash F.
Sent: 17 Nopember 2008 17:50
To: [email protected]
Subject: Re: strange ‘too many open files’ errors with nginx 0.6.32

oh, and one more thing, restarting nginx seems to fix everything (until
the
error pops up again a day or so later :slight_smile:

On Mon, Nov 17, 2008 at 2:48 AM, Arash F. [email protected]
wrote:

hi,

i’m seeing a strange ‘too many open files’ error with nginx (which I’m
sure
many of you have seen before):

2008/11/17 02:15:02 [alert] 21748#0: accept() failed (24: Too many open
files) while accepting new connection on 67.228.78.114:80

strangely enough, I have the following line in /etc/security/limits:

  • hard nofile 524288

and ulimit -n returns the same number. I have 2 workers set to a max of
262144 connections. upon receiving that error (and after past
occurrences)
I’ve run lsof and gotten < 10k sockets. additionally, I have a nearly
identically configured machine (running twisted) where each process is
handling ~15k connections. so I’m a loss as to what’s going on here (and
suspect some subtle bug in nginx or in my server configuration. any help
would be greatly appreciated!


Arash F.
CTO, Dropbox
913.707.5875 (m)