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

[Tao Ma]

Hi guys,I got a problem.

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 system.When it was killed,the following log was found in /var/log/messages("Dnsserv" is the hostname):
Sep  8 00:21:02 Dnsserv Out of memory: kill process 23999 (unbound) score 125734 or a child
Sep  8 00:21:02 Dnsserv Killed process 23999 (unbound)


I only specify some normal items in unbound.conf and it was shown bellow:
server:
        verbosity: 1
        statistics-interval: 1
        num-threads: 1
        interface: 0.0.0.0
        interface-automatic: no
        port: 53
        msg-cache-size: 4m
        rrset-cache-size: 4m
        do-ip4: yes
        do-ip6: no
        do-udp: yes
        access-control: 10.0.0.0/8 allow
        access-control: 127.0.0.0/8 allow
        statistics-interval: 10
        #chroot: "/var/unbound"
        username: "unbound"
        directory: "/var/unbound"
        logfile: ""
        use-syslog: no
        log-time-ascii: yes
        pidfile: "/var/run/unbound.pid"
        root-hints: "/var/unbound/named.cache"
        hide-version: no
        version: "hiahia"

local-zone: "mydns123.com." static
        local-data: "mydns123.com. SOA ns.mydns123.com tj.mydns123.com. 1 28800 7200 604800 86400"
        local-data: "mydns123.com. IN NS ns.mydns123.com."
        local-data: "mydns123.com. IN NS ns2.mydns123.com."
        local-data: "ns.mydns123.com. IN A 192.1.1.1"
        local-data: "ns1.mydns123.com. IN A 192.1.1.2"
        local-data: "www.mydns123.com. IN A 192.1.1.3"
        local-data: "ftp.mydns123.com. IN A 192.1.1.4"
        local-data: "wa.mydns123.com. IN A 192.1.1.5"
        local-data: "pp.mydns123.com. IN CNAME wa.mydns123.com."

        local-zone:"1.1.192.in-addr.arpa" static
        local-data: "mydns123.com. SOA ns.mydns123.com tj.mydns123.com. 1 28800 7200 604800 86400"
        local-data: "mydns123.com. IN NS ns.mydns123.com."
        local-data: "mydns123.com. IN NS ns2.mydns123.com."
        local-data:"1.1.1.192.in-addr.arpa IN PTR ns.mydns123.com."
        local-data:"2.1.1.192.in-addr.arpa IN PTR ns1.mydns123.com."
        local-data:"3.1.1.192.in-addr.arpa IN PTR www.mydns123.com."
        local-data:"4.1.1.192.in-addr.arpa IN PTR ftp.mydns123.com."
        local-data:"5.1.1.192.in-addr.arpa IN PTR was.mydns123.com."

Is there any mistake in my configuration?Or is it a bug of unbound?




___________________________________________________________
奜玩莺卡等䜠发邮箱莺卡党新䞊线
http://card.mail.cn.yahoo.com/

[Paul Wouters]

Post by Tao Ma
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
        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

[Tao Ma]

Post by Tao Ma

Post by Tao Ma
I installed unbound 1.3.3 and found when the

unbound is running,it use

Post by Tao Ma
more and more memory..
I even found it occupied 96% of memory and then

was killed by the

Post by Tao Ma
Sep 8 00:21:02 Dnsserv Out of memory: kill process

23999 (unbound) score

Post by Tao Ma
I only specify some normal items in unbound.conf
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-***@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
install".

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
(0xb7f45000)
libcrypto.so.0.9.8 => /usr/lib/libcrypto.so.0.9.8
(0xb7e0f000)
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?
Thanks.


___________________________________________________________
好玩贺卡等你发，邮箱贺卡全新上线！
http://card.mail.cn.yahoo.com/

[W.C.A. Wijngaards]

Hi Tao Ma,

The memory statistics that unbound makes can help?

It prints them into the log file (depending on verbosity
level and --enable-debug to configure) like this:
debug: cache memory msg=1620127 rrset=2964258 infra=738306 val=128832

Also the memory statistics are printed when you enable extended
statistics and do unbound-control stats, like this:
mem.cache.rrset=2980222
mem.cache.message=1633031
mem.mod.iterator=16460
mem.mod.validator=130126

Best regards,
Wouter

Post by Tao Ma

Post by Tao Ma

Post by Tao Ma
I installed unbound 1.3.3 and found when the

unbound is running,it use

Post by Tao Ma
more and more memory..
I even found it occupied 96% of memory and then

was killed by the

Post by Tao Ma
Sep 8 00:21:02 Dnsserv Out of memory: kill process

23999 (unbound) score

Post by Tao Ma
I only specify some normal items in unbound.conf
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
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
install".
linux-gate.so.1 => (0xffffe000)
libssl.so.0.9.8 => /usr/lib/libssl.so.0.9.8
(0xb7f45000)
libcrypto.so.0.9.8 => /usr/lib/libcrypto.so.0.9.8
(0xb7e0f000)
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?
Thanks.
___________________________________________________________
好玩贺卡等你发，邮箱贺卡全新上线！
http://card.mail.cn.yahoo.com/
_______________________________________________
Unbound-users mailing list
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users

[Tao Ma]

Hi Wouter,thanks very much for your reply.

I tried the method you mentioned,here is the info I got from the log:
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
This info keeps the same from the start of the unbound(I have only about 5 records in unbound).

I also found the value of mem.total.sbrk printed by "unbound-control stats" keeps growing.The item "mem.total.sbrk" is an estimate of the heap size of unbound in bytes,it is close to the memory used by unbound.

I have a script keeps recording mem.total.sbrk in a file,and found when mem.total.sbrk grew to 2017873920,unbound was killed by system(My server has a 2G memory).

Unbound runs in a Gentoo Linux with 2.6.23 kernel,does the problem relate to the OS?

Thanks & Best regards,
Tao Ma



--- 09幎9月10日呚四, W.C.A. Wijngaards <***@NLnetLabs.nl> 写道

发件人: W.C.A. Wijngaards <***@NLnetLabs.nl>
䞻题: Re: [Unbound-users] 回倍 Re: Is there any memory leak in unbound?
收件人: "Tao Ma" <***@yahoo.com.cn>
抄送: "Paul Wouters" <***@xelerance.com>, unbound-***@unbound.net
日期: 2009幎9月10日,å‘šå››,䞋午10:25

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Tao Ma,

The memory statistics that unbound makes can help?

It prints them into the log file (depending on verbosity
level and --enable-debug to configure) like this:
debug: cache memory msg=1620127 rrset=2964258 infra=738306 val=128832

Also the memory statistics are printed when you enable extended
statistics and do unbound-control stats, like this:
mem.cache.rrset=2980222
mem.cache.message=1633031
mem.mod.iterator=16460
mem.mod.validator=130126

Best regards,
   Wouter




___________________________________________________________
奜玩莺卡等䜠发邮箱莺卡党新䞊线
http://card.mail.cn.yahoo.com/

[Tao Ma]

What's more,I found that the memory occupied by unbound would not be freed when all requets finished,that means,unbound only consume memory but never free them.
What does unbound hold the memory for?It quite like that  a memory leak happens...
Anyone else has been in the same trouble?


Best Regards,
Tao Ma


--- 09幎9月11日呚五, Tao Ma <***@yahoo.com.cn> 写道

发件人: Tao Ma <***@yahoo.com.cn>
䞻题: Re: [Unbound-users] Is there any memory leak in unbound?
收件人: "W.C.A. Wijngaards" <***@nlnetlabs.nl>
抄送: unbound-***@unbound.net
日期: 2009幎9月11日,呚五,䞊午11:50

Hi Wouter,thanks very much for your reply.

I tried the method you mentioned,here is the info I got from the log:
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
This info keeps the same from the start of the unbound(I have only about 5 records in unbound).

I also found the value of mem.total.sbrk printed by "unbound-control stats" keeps growing.The item "mem.total.sbrk" is an estimate of the heap size of unbound in bytes,it is close to the memory used by unbound.

I have a script keeps recording mem.total.sbrk in a file,and found when mem.total.sbrk grew to 2017873920,unbound was killed by system(My server has a 2G memory).

Unbound runs in a Gentoo Linux with 2.6.23 kernel,does the problem relate to the OS?

Thanks & Best regards,
Tao Ma



---
09幎9月10日呚四, W.C.A. Wijngaards <***@NLnetLabs.nl> 写道

发件人: W.C.A. Wijngaards <***@NLnetLabs.nl>
䞻题: Re: [Unbound-users] 回倍 Re: Is there any memory leak in unbound?
收件人: "Tao Ma" <***@yahoo.com.cn>
抄送: "Paul Wouters" <***@xelerance.com>, unbound-***@unbound.net
日期: 2009幎9月10日,å‘šå››,䞋午10:25

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hi Tao Ma,

The memory statistics that unbound makes can help?

It prints them into the log file (depending on verbosity
level and --enable-debug to configure) like this:
debug: cache memory msg=1620127 rrset=2964258 infra=738306 val=128832

Also the memory statistics are printed when you enable extended
statistics and do
unbound-control stats, like this:
mem.cache.rrset=2980222
mem.cache.message=1633031
mem.mod.iterator=16460
mem.mod.validator=130126

Best regards,
   Wouter





奜玩莺卡等䜠发邮箱莺卡党新䞊线
-----䞋面䞺附件内容-----

_______________________________________________
Unbound-users mailing list
Unbound-***@unbound.net
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users


___________________________________________________________
奜玩莺卡等䜠发邮箱莺卡党新䞊线
http://card.mail.cn.yahoo.com/

[Artis Caune]

Post by Tao Ma
What's more,I found that the memory occupied by unbound would not be freed when all requets finished,that means,unbound only consume memory but never free them.
What does unbound hold the memory for?It quite like that  a memory leak happens...
Anyone else has been in the same trouble?

Hi,

do you reload unbound while it's running ?




--
Artis Caune

   Everything should be made as simple as possible, but not simpler.

[W.C.A. Wijngaards]

Hi Tao Ma,

Post by Tao Ma
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888

The values from your log file look very reasonable. No problems here.
The sbrk keeps growing?

Ok, it is possible to enable even more memory debug, configure with
- --enable-alloc-checks and compile again. Unbound prints:
info: Memory conditions: %u front=%u back=%u mesh=%u msg=%u rrset=%u
infra=%u iter=%u val=%u alloccache=%u globalalloccache=%u me=%u
But with values.

It also prints a log line for *every* alloc call. So, you can see where
in the code it allocates all that memory. This will make the
server very slow. If you send me the log file (off-list to
***@nlnetlabs.nl , then I can parse it with testcode/memstats.c;
or you can do that yourself: make memstats ; ./memstats < logfile ).

To answer your question, unbound keeps some memory around, up to a
maximum, because free-ing something that you need later again is slow.
This should top out at 40Mb for the standard config.

Gentoo? Is it 64bit? Does it use a 'special' memory allocator in libc?

Best regards,
Wouter

Post by Tao Ma
What's more,I found that the memory occupied by unbound would not be
freed when all requets finished,that means,unbound only consume memory
but never free them.
What does unbound hold the memory for?It quite like that a memory leak
happens...
Anyone else has been in the same trouble?
Best Regards,
Tao Ma
主题: Re: [Unbound-users] Is there any memory leak in unbound?
日期: 2009年9月11日,周五,上午11:50
Hi Wouter,thanks very much for your reply.
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
This info keeps the same from the start of the unbound(I have only
about 5 records in unbound).
I also found the value of mem.total.sbrk printed by "unbound-control
stats" keeps growing.The item "mem.total.sbrk" is an estimate of the
heap size of unbound in bytes,it is close to the memory used by unbound.
I have a script keeps recording mem.total.sbrk in a file,and found
when mem.total.sbrk grew to 2017873920,unbound was killed by
system(My server has a 2G memory).
Unbound runs in a Gentoo Linux with 2.6.23 kernel,does the problem
relate to the OS?
Thanks & Best regards,
Tao Ma
写道：
主题: Re: [Unbound-users] 回复： Re: Is there any memory leak in
unbound?
日期: 2009年9月10日,周四,下午10:25
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi Tao Ma,
The memory statistics that unbound makes can help?
It prints them into the log file (depending on verbosity
debug: cache memory msg=1620127 rrset=2964258 infra=738306 val=128832
Also the memory statistics are printed when you enable extended
mem.cache.rrset=2980222
mem.cache.message=1633031
mem.mod.iterator=16460
mem.mod.validator=130126
Best regards,
Wouter
------------------------------------------------------------------------
好玩贺卡等你发，邮箱贺卡全新上线！
<http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>
-----下面为附件内容-----
_______________________________________________
Unbound-users mailing list
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users
------------------------------------------------------------------------
好玩贺卡等你发，邮箱贺卡全新上线！
<http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>

[W.C.A. Wijngaards]

Hi Tao Ma,

Just realized something else, the sbrk estimate is a dirty hack.
Gentoo may report a different sbrk() value, and the dirty hack
does not work. The sbrk() value keeps growing but maybe it is
not true that unbound's memory keeps growing. Does the 'top' or
'ps' memory number also grow?

Best regards,
Wouter

Post by W.C.A. Wijngaards
Hi Tao Ma,

Post by Tao Ma
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888

The values from your log file look very reasonable. No problems here.
The sbrk keeps growing?
Ok, it is possible to enable even more memory debug, configure with
info: Memory conditions: %u front=%u back=%u mesh=%u msg=%u rrset=%u
infra=%u iter=%u val=%u alloccache=%u globalalloccache=%u me=%u
But with values.
It also prints a log line for *every* alloc call. So, you can see where
in the code it allocates all that memory. This will make the
server very slow. If you send me the log file (off-list to
or you can do that yourself: make memstats ; ./memstats < logfile ).
To answer your question, unbound keeps some memory around, up to a
maximum, because free-ing something that you need later again is slow.
This should top out at 40Mb for the standard config.
Gentoo? Is it 64bit? Does it use a 'special' memory allocator in libc?
Best regards,
Wouter

Post by Tao Ma
What's more,I found that the memory occupied by unbound would not be
freed when all requets finished,that means,unbound only consume memory
but never free them.
What does unbound hold the memory for?It quite like that a memory leak
happens...
Anyone else has been in the same trouble?
Best Regards,
Tao Ma
;˜: Re: [Unbound-users] Is there any memory leak in unbound?
å: 2009t911å,h”,

H11:50

Post by Tao Ma
Hi Wouter,thanks very much for your reply.
debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
This info keeps the same from the start of the unbound(I have only
about 5 records in unbound).
I also found the value of mem.total.sbrk printed by "unbound-control
stats" keeps growing.The item "mem.total.sbrk" is an estimate of the
heap size of unbound in bytes,it is close to the memory used by unbound.
I have a script keeps recording mem.total.sbrk in a file,and found
when mem.total.sbrk grew to 2017873920,unbound was killed by
system(My server has a 2G memory).
Unbound runs in a Gentoo Linux with 2.6.23 kernel,does the problem
relate to the OS?
Thanks & Best regards,
Tao Ma
™S
;˜: Re: [Unbound-users] Þ  Re: Is there any memory leak in
unbound?
å: 2009t910å,hÛ,H10:25
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Hi Tao Ma,
The memory statistics that unbound makes can help?
It prints them into the log file (depending on verbosity
debug: cache memory msg=1620127 rrset=2964258 infra=738306 val=128832
Also the memory statistics are printed when you enable extended
mem.cache.rrset=2980222
mem.cache.message=1633031
mem.mod.iterator=16460
mem.mod.validator=130126
Best regards,
Wouter
------------------------------------------------------------------------
}©:aI`Ñ®±:ah°

¿

Post by Tao Ma
<http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>
-----b:Dö…¹-----
_______________________________________________
Unbound-users mailing list
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users
------------------------------------------------------------------------
}©:aI`Ñ®±:ah°

¿

Post by Tao Ma
<http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>

_______________________________________________
Unbound-users mailing list
Unbound-***@unbound.net
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users

[Tao Ma]

Hi,I didn't reload unbound while it's running,I always start it and then send requests to the server,then it will consume more and more memory.

--- 09幎9月11日呚五, Artis Caune <***@gmail.com> 写道

发件人: Artis Caune <***@gmail.com>
䞻题: Re: [Unbound-users] Is there any memory leak in unbound?
收件人: "Tao Ma" <***@yahoo.com.cn>
抄送: ***@nlnetlabs.nl, unbound-***@unbound.net
日期: 2009幎9月11日,呚五,䞋午7:02

Hi,

do you reload unbound while it's running ?




--
Artis Caune

聜 聜Everything should be made as simple as possible, but not simpler.



___________________________________________________________
奜玩莺卡等䜠发邮箱莺卡党新䞊线
http://card.mail.cn.yahoo.com/

[Tao Ma]

Hi Wouter,I'm very glad to tell you,I patched the diff file you gave me and now unbound works well on the server,thanks very much!

I'm sorry that I didn't tell you clearly about the compile problem.When we found we couldn't compile the code with "--enable-alloc-checks",we added another option "--enable-static-exe" and then finished compiling smoothly,that's all.

Thanks again for your warmhearted help.

Best Regards,
Tao Ma

--- 09幎9月16日呚䞉, W.C.A. Wijngaards <***@nlnetlabs.nl> 写道

发件人: W.C.A. Wijngaards <***@nlnetlabs.nl>
䞻题: Re: [Unbound-users] Is there any memory leak in unbound?
收件人: "Tao Ma" <***@yahoo.com.cn>
日期: 2009幎9月16日,呚䞉,䞋午8:47

Hi Tao Ma,

Does this solve your memory leak problem?

Index: libunbound/libworker.c
===================================================================
--- libunbound/libworker.c    (revision 1831)
+++ libunbound/libworker.c    (working copy)
@@ -514,6 +514,7 @@
      ldns_buffer_write_u16_at(w->back->udp_buff, 2, qflags);
      if(local_zones_answer(ctx->local_zones, &qinfo, &edns,
          w->back->udp_buff, w->env->scratch)) {
+        regional_free_all(w->env->scratch);
          libworker_fillup_fg(q, LDNS_RCODE_NOERROR,
              w->back->udp_buff, sec_status_insecure);
          libworker_delete(w);
@@ -630,6 +631,7 @@
      ldns_buffer_write_u16_at(w->back->udp_buff, 2, qflags);
      if(local_zones_answer(w->ctx->local_zones, &qinfo, &edns,
          w->back->udp_buff, w->env->scratch)) {
+        regional_free_all(w->env->scratch);
          q->msg_security = sec_status_insecure;
          add_bg_result(w, q, w->back->udp_buff, UB_NOERROR);
          free(qinfo.qname);
Index: daemon/worker.c
===================================================================
--- daemon/worker.c    (revision 1831)
+++ daemon/worker.c    (working copy)
@@ -806,6 +806,7 @@
      }
      if(local_zones_answer(worker->daemon->local_zones, &qinfo, &edns,
          c->buffer, worker->scratchpad)) {
+        regional_free_all(worker->scratchpad);
          if(ldns_buffer_limit(c->buffer) == 0) {
              comm_point_drop_reply(repinfo);
              return 0;

Sorry about the compile problem with unbound-host and alloc-checks, I'll
look at it later.

Best regards,
    Wouter

Hi Wouter,thanks for your reply.
We use a 32-bit Gentoo and I don't think there's a special memory
allocator in libc.
I can always find the memory consumed by unbound keep growing from ouput
of 'ps' and 'top',the result of sbrk estimate seems to be in accord with
this.
Following your suggestion,we configure with --enable-alloc-checks and
compile unbound again,now we can find the "Memory conditions" value keep
growing in the log.
Reading /var/unbound/log_test of size 38767531
1% 2% 3% 4% 5% 6% 7% 8% 9% 10% 11% 12% 13% 14% 15% 16% 17% 18% 19% 20%
21% 22% 23% 24% 25% 26% 27% 28% 29% 30% 31% 32% 33% 34% 35% 36% 37% 38%
39% 40% 41% 42% 43% 44% 45% 46% 47% 48% 49% 50% 51% 52% 53% 54% 55% 56%
57% 58% 59% 60% 61% 62% 63% 64% 65% 66% 67% 68% 69% 70% 71% 72% 73% 74%
75% 76% 77% 78% 79% 80% 81% 82% 83% 84% 85% 86% 87% 88% 89% 90% 91% 92%
93% 94% 95% 96% 97% 98% 99% 100% done
------------
0 / 0 total in 0 code lines
We got a complete record in the log from the start of unbound to the
exiting of unbound,and the log file has been attatched to the mail,maybe
it will help you to analyse.
Let me know if you find any mistakes in my operation.
By the way,when we configure unbound-1.3.3 with "./configure
--prefix=/usr/local/unbound --enable-alloc-checks" and compile,we got
/home/mt/tools/unbound-1.3.3/smallapp/unbound-host.c:233: undefined
reference to `unbound_stat_malloc_log'
/home/mt/tools/unbound-1.3.3/smallapp/unbound-host.c:255: undefined
reference to `unbound_stat_free_log'
/home/mt/tools/unbound-1.3.3/smallapp/unbound-host.c:247: undefined
reference to `unbound_stat_free_log'
/home/mt/tools/unbound-1.3.3/smallapp/unbound-host.c:384: undefined
reference to `unbound_stat_free_log'
collect2: ld returned 1 exit status
make: *** [unbound-host] Error 1
We did some modification to the code and finished compiling.Is there
anything wrong in the code?
Thanks & Best regards,
Tao Ma
     䞻题: Re: [Unbound-users] Is there any memory leak in unbound?
     日期: 2009幎9月15日,呚二,䞋午5:29
     -----BEGIN PGP SIGNED MESSAGE-----
     Hash: SHA1
     Hi Tao Ma,
     Just realized something else, the sbrk estimate is a dirty hack.
     Gentoo may report a different sbrk() value, and the dirty hack
     does not work. The sbrk() value keeps growing but maybe it is
     not true that unbound's memory keeps growing. Does the 'top' or
     'ps' memory number also grow?
     Best regards,
     Wouter
      > Hi Tao Ma,
      >
      >> debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
      >
      > The values from your log file look very reasonable. No problems here.
      > The sbrk keeps growing?
      >
      > Ok, it is possible to enable even more memory debug, configure with
      > info: Memory conditions: %u front=%u back=%u mesh=%u msg=%u rrset=%u
      > infra=%u iter=%u val=%u alloccache=%u globalalloccache=%u me=%u
      > But with values.
      >
      > It also prints a log line for *every* alloc call. So, you can see
     where
      > in the code it allocates all that memory. This will make the
      > server very slow. If you send me the log file (off-list to
     can parse it with testcode/memstats.c;
      > or you can do that yourself: make memstats ; ./memstats < logfile ).
      >
      > To answer your question, unbound keeps some memory around, up to a
      > maximum, because free-ing something that you need later again is
     slow.
      > This should top out at 40Mb for the standard config.
      >
      > Gentoo? Is it 64bit? Does it use a 'special' memory allocator in
     libc?
      >
      > Best regards,
      > Wouter
      >
      >> What's more,I found that the memory occupied by unbound would not be
      >> freed when all requets finished,that means,unbound only consume
     memory
      >> but never free them.
      >> What does unbound hold the memory for?It quite like that a
     memory leak
      >> happens...
      >> Anyone else has been in the same trouble?
      >
      >
      >> Best Regards,
      >> Tao Ma
      >
      >
      >
      >
      >> ;聵: Re: [Unbound-users] Is there any memory leak in unbound?
      >> 氓: 2009t911氓,h聰,
      > H11:50
      >
      >> Hi Wouter,thanks very much for your reply.
      >
      >> I tried the method you mentioned,here is the info I got from the
      >> debug: cache memory msg=16528 rrset=33040 infra=672 val=24888
      >> This info keeps the same from the start of the unbound(I have only
      >> about 5 records in unbound).
      >
      >> I also found the value of mem..total.sbrk printed by
     "unbound-control
      >> stats" keeps growing.The item "mem.total.sbrk" is an estimate of the
      >> heap size of unbound in bytes,it is close to the memory used by
     unbound.
      >
      >> I have a script keeps recording mem.total.sbrk in a file,and found
      >> when mem.total.sbrk grew to 2017873920,unbound was killed by
      >> system(My server has a 2G memory).
      >
      >> Unbound runs in a Gentoo Linux with 2.6.23 kernel,does the problem
      >> relate to the OS?
      >
      >> Thanks & Best regards,
      >> Tao Ma
      >
      >
      >
      >> 聶S
      >
      >
      >> ;聵: Re: [Unbound-users] è„Œ Re: Is there any memory leak in
      >> unbound?
      >> 氓: 2009t910氓,h脹,H10:25
      >
      >> -----BEGIN PGP SIGNED MESSAGE-----
      >> Hash: SHA1
      >
      >> Hi Tao Ma,
      >
      >> The memory statistics that unbound makes can help?
      >
      >> It prints them into the log file (depending on verbosity
      >> debug: cache memory msg=1620127 rrset=2964258 infra=738306
      >> val=128832
      >
      >> Also the memory statistics are printed when you enable extended
      >> mem.cache.rrset=2980222
      >> mem.cache.message=1633031
      >> mem.mod.iterator=16460
      >> mem.mod.validator=130126
      >
      >> Best regards,
      >> Wouter
      >
      >
      >>
     ------------------------------------------------------------------------
      >> }挏:aI`脩庐區:ah掳
      > 驎
      >>
     <http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>
      >
      >> -----b:D枚聟鹿-----
      >
      >> _______________________________________________
      >> Unbound-users mailing list
      >> http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users
      >
      >
      >>
     ------------------------------------------------------------------------
      >> }挏:aI`脩庐區:ah掳
      > 驎
      >>
     <http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>
      >
      >
     _______________________________________________
     Unbound-users mailing list
     http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users
     -----BEGIN PGP SIGNATURE-----
     Version: GnuPG v1.4.9 (GNU/Linux)
     Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org/
     iEYEARECAAYFAkqvXmkACgkQkDLqNwOhpPisoACeKatx2Pl4/B7k5MtJ86KTPBaR
     2WkAn0sd6bW9UDjCbdQ4pOLHW2WiFOg+
     =NrkI
     -----END PGP SIGNATURE-----
------------------------------------------------------------------------
奜玩莺卡等䜠发邮箱莺卡党新䞊线
<http://cn.rd.yahoo.com/mail_cn/tagline/card/*http://card.mail.cn.yahoo.com/>

___________________________________________________________
奜玩莺卡等䜠发邮箱莺卡党新䞊线
http://card.mail.cn.yahoo.com/