[Date Prev][Date Next]
[Chronological]
[Thread]
[Top]
Re: How to raise the open file descriptor limit for slapd?
- To: David Landgren <david@landgren.net>
- Subject: Re: How to raise the open file descriptor limit for slapd?
- From: matthew sporleder <msporleder@gmail.com>
- Date: Tue, 4 Oct 2005 14:40:15 -0400
- Cc: openldap-software@OpenLDAP.org
- Content-disposition: inline
- Domainkey-signature: a=rsa-sha1; q=dns; c=nofws; s=beta; d=gmail.com; h=received:message-id:date:from:reply-to:to:subject:cc:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=dtmBeW9LqRwraLK4V34Q2xWpYMfrWeiGAfncFswWyu4NitD5W4UbE4bduhrK4e48nHgm28gAzth2NgKRXjn6CerSJMXbxcQ3xthIdog3sOnMgFVPwSbJrWDJGrGoBXpGXI4XQqAvSPwA+ypCHulfk07Nu2vvPrWxK2ms9U+Aj+A=
- In-reply-to: <43424FC8.5010706@landgren.net>
- References: <43424FC8.5010706@landgren.net>
You should be able to set ulimit values in your daemon's init script,
or set them system-wide.
_Matt
On 10/4/05, David Landgren <david@landgren.net> wrote:
> List,
>
> I have a fairly heavily used openldap-2.2.28 server in production and I
> am running into the 1024 fd descriptor limit. From time to time I get
>
> 1024 beyond descriptor table limit 1024
>
> on my syslog host. I found the following message in the archives:
>
> http://www.openldap.org/lists/openldap-software/200303/msg00858.html
> beyond descriptor table size error
>
> One of the suggestions was to keep an eye on open sockets. So I wrote
> the following script:
>
> netstat --inet --numeric-hosts --numeric-ports --numeric-users \
> | tail +3 \
> | awk '{print $6}' \
> | sort \
> | uniq -c \
> | awk '{n+=$1; print} END {printf("%7d\n", n)}'
>
> which I run every five minutes. Within minutes of restarting slapd this
> morning, I was seeing:
>
> Tue Oct 4 09:09:02 CEST 2005
> 40 CLOSE_WAIT
> 242 ESTABLISHED
> 60 TIME_WAIT
> 342
> Tue Oct 4 09:14:03 CEST 2005
> 37 CLOSE_WAIT
> 293 ESTABLISHED
> 65 TIME_WAIT
> 395
> Tue Oct 4 09:19:03 CEST 2005
> 46 CLOSE_WAIT
> 335 ESTABLISHED
> 37 TIME_WAIT
> 418
>
> Now during the heavy activity during the morning I see
>
> Tue Oct 4 11:29:11 CEST 2005
> 131 CLOSE_WAIT
> 664 ESTABLISHED
> 2 FIN_WAIT1
> 1 FIN_WAIT2
> 32 TIME_WAIT
> 830
> Tue Oct 4 11:34:12 CEST 2005
> 130 CLOSE_WAIT
> 647 ESTABLISHED
> 2 FIN_WAIT1
> 14 TIME_WAIT
> 793
> Tue Oct 4 11:39:12 CEST 2005
> 131 CLOSE_WAIT
> 660 ESTABLISHED
> 2 FIN_WAIT1
> 1 FIN_WAIT2
> 40 TIME_WAIT
> 834
> Tue Oct 4 11:44:13 CEST 2005
> 130 CLOSE_WAIT
> 659 ESTABLISHED
> 2 FIN_WAIT1
> 17 TIME_WAIT
> 808
>
>
> So the close_wait values are getting higher, and I may be able to tune
> this down by clamping the idle_timeout setting, but as it is, I don't
> think I have enough headroom.
>
> One of the suggestions was to increase FD_SETSIZE, which seems to me to
> be the right thing to do, but I'm not quite sure what the recommended
> practice is. In the CHANGES file I see a mention of OPENLDAP_FD_SETSIZE
> but a quick grep through the source does not turn up any other reference.
>
> Do I define this as an environment variable, or should I define
> CFLAGS=-DOPENLDAP_FD_SETSIZE=2048 , or something else?
>
> Your suggestions are appreciated,
> David
> --
> "It's overkill of course, but you can never have too much overkill."
>