While going though my logs I found that BIND9 on Debian version 5.x is giving a warning which read as follows:
max open files (1024) is smaller than max sockets (4096)
How do I fix this problem?
The “open files is smaller than max sockets” problem is kernel bug which is already fixed in 2.6.28. Linux kernel returns EPERM when RLIMIT_NOFILE is set to RLIM_INFINITY. To fix this issue update your Linux kernel using yum or apt-get command.
# yum update
OR
# apt-get update
# apt-get upgrade
Reboot the server. Verify error is not reported after boot:
# tail -f /var/log/messages
9 comment
Soultion? ;-)
Thanks for the heads up.
how to fix on debian lenny 5.0 ( 2.6.26-2-686):
# vim /etc/default/bind9
add “-S 1024” to “OPTIONS” line:
restart bind:
# /etc/init.d/bind9 restart
check for errors:
# cat /var/log/syslog
ooops :)
# vim /etc/default/bind9
add “-S 1024″ to “OPTIONS†line:
# run resolvconf?
RESOLVCONF=yes
# startup options for the server
OPTIONS="-u bind -S 1024"
how to fix the error in freebsd?
sysctl kern.maxfilesperproc=4096
There’s another way of fixing it on Linux:
At the beginning of the “start” section of /etc/init.d/bind9 add i.e.:
ulimit -HSn 8192
so you have some extra, spare file descriptors for the whole process,
and then restart the bind using:
# /etc/init.d/bind9 restart
and voila, log reads:
[…]
named[19060]: using up to 4096 sockets
[…]
The preceding “ulimit -HSn 8192” command might better go in /etc/default/bind9 on debian to keep package updates (ie. changed config files) a little smoother.
If you are using Centos or redhat, edit /etc/sysconfig/named and at the bottom, at,
OPTIONS=”-4 -S 1024″
It think the author’s claim of upgrading the kernel it’s false and can be dangers. I got a vmware application runnning on the host, upgrading the kernel can simply break it