[Unbound-users] 回复: Re: Is there any memory leak in unbound?

Tao Ma matao_reg at yahoo.com.cn
Wed Sep 9 07:01:23 UTC 2009

--- Paul Wouters <paul at xelerance.com>写道:

> On Tue, 8 Sep 2009, Tao Ma wrote:
> > I installed unbound 1.3.3 and found when the
> unbound is running,it use
> > more and more memory..
> > I even found it occupied 96% of memory and then
> was killed by the
> > Sep 8 00:21:02 Dnsserv Out of memory: kill process
> 23999 (unbound) score
> > I only specify some normal items in unbound.conf
> and it was shown bellow:
> > server:
> >  msg-cache-size: 4m
> >  rrset-cache-size: 4m
> Is this an embedded device with little memory? eg
> 16mb or 32mb?
> The OOM killer might be killing the biggest memory
> consumer, which
> not neccessarilly means that process is the one
> leaking memory. 
> unbound might be running stable at just over 8mb,
> and some other
> process might be slowly allocating more memory over
> time, but still
> end up using a smaller memory footprint, for which
> unbound takes the
> blame (and the shot)
> Paul

Hi Paul,thanks for your reply.
Sorry to send this mail again because I forgot to cc
to unbound-users at unbound.net in the last mail.

It's my fault to forget to tell you my server's
hardware and software enviroment in the first mail.
The server has 2G memory and the CPU is Intel(R)
Celeron(R) M processor 1.50GHz.I didn't specify any
other options except "--prefix=/usr/local/unboud" in
configure,then installed unbound by "make&&make

Here is the shared library used by unboud:
linux-gate.so.1 =>  (0xffffe000)
libssl.so.0.9.8 => /usr/lib/libssl.so.0.9.8
libcrypto.so.0.9.8 => /usr/lib/libcrypto.so.0.9.8
libpthread.so.0 => /lib/libpthread.so.0 (0xb7df8000)
libc.so.6 => /lib/libc.so.6 (0xb7cd0000)
libdl.so.2 => /lib/libdl.so.2 (0xb7ccc000)
/lib/ld-linux.so.2 (0xb7f8d000)

When processing lots of DNS requests(like about 50,000
per second),unbound was found consume more and more
memory.I found it by reading the output of command
"top".Unobund wouldn't stop consuming more memory
until killed by OOM killer.
In my opinion,the memory consumed by unbound should be
stable when the count of requests per
second(say,3,000) is stable,is it right?

Any hints for me to find the reason of the problem?


More information about the Unbound-users mailing list