[Unbound-users] Unbound v1.4.4 in forked mode load unbalanced

Slingerland, Michael van Michael.van.Slingerland at t-mobile.nl
Tue May 25 14:18:13 UTC 2010


Hi,

I am planning to use Unbound v1.4.4 on OpenBSD 4.5 (i386) to replace my current bind9 instances.

Currently I am running it as test on one of my low traffic dns servers ( about 1k request per second).

But what I notice is that 1 process is serving almost all traffic while the other process is handling only a fraction of the requests.

The compile options I used are:
./configure --prefix=/opt/unbound-1.4.4 \
--with-ssl=/usr \
--with-libevent=/usr \
--without-pthreads \
--with-chroot-dir=/var/unbound \
--with-pidfile=/var/run/unbound.pid \
--with-conf-file=/var/unbound/etc/unbound.conf \
--with-username=named

Mofified parameters in Unbound.conf
verbosity: 1
extended-statistics: yes
num-threads: 2
outgoing-range: 1024
msg-cache-size: 300m
num-queries-per-thread: 1024
rrset-cache-size: 600m
do-ip6: no
access-control: 0.0.0.0/0 allow
prefetch: yes
prefetch-key: yes
module-config: "iterator"



Here is an output of the stats
# /opt/unbound-1.4.4/sbin/unbound-control stats
thread0.num.queries=44189
thread0.num.cachehits=33671
thread0.num.cachemiss=10518
thread0.num.prefetch=1371
thread0.num.recursivereplies=10518

thread1.num.queries=2361021
thread1.num.cachehits=2203077
thread1.num.cachemiss=157944
thread1.num.prefetch=24706
thread1.num.recursivereplies=157935

total.num.queries=2405210
total.num.cachehits=2236748
total.num.cachemiss=168462
total.num.prefetch=26077
total.num.recursivereplies=168453

I do not expect a perfect 50/50 load distribution, but this is totally unbalanced.

Does anyone have a clue what is happening here? And how I can solve this.
I really would like to use unbound in forked mode.....

Thanks,
Michael

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.nlnetlabs.nl/pipermail/unbound-users/attachments/20100525/5c0044c9/attachment.htm>


More information about the Unbound-users mailing list