[Unbound-users] about memory leak? Unbound 1.4.22

Discussion:

Techs_Maru

2015-05-22 12:36:56 UTC

Hi,Folks,

I have questions about Unbound 1.4.22

I began to use recently Unbound.
before that i was using ISC-BIND.

Unbound to service in a production environment,
It seems like memory leak as the recursive query increases.
Is high and utilization compared to ISC-BIND.
(but,Unbound and ISC-BIND rr-cache-size configure parameter is different.)

Unbound is easy to memory leak?
Or ,Unbound bugs?
Or , It is proportional to the cache size value ?

Now "Resident set size" is about 4GB.

It does parameter tuning is bad?.....

Best regards

l***@kwsoft.de

2015-05-22 13:49:51 UTC

Permalink

Post by Techs_Maru
Hi,Folks,
I have questions about Unbound 1.4.22
I began to use recently Unbound.
before that i was using ISC-BIND.
Unbound to service in a production environment,
It seems like memory leak as the recursive query increases.
Is high and utilization compared to ISC-BIND.
(but,Unbound and ISC-BIND rr-cache-size configure parameter is different.)
Unbound is easy to memory leak?
Or ,Unbound bugs?
Or , It is proportional to the cache size value ?
Now "Resident set size" is about 4GB.
It does parameter tuning is bad?.....

Hello,

have you read http://www.unbound.net/documentation/howto_optimise.html ??

"Due to malloc overhead, the total memory usage is likely to rise to
double (or 2.5x) the total cache memory that is entered into the
config."

And no, memory leaks are not common in Unbound to my knowledge at least.

Regards

Andreas

Techs_Maru

2015-05-26 02:40:31 UTC

Permalink

Hi,Andreas,

thank you for reply.

Post by l***@kwsoft.de
have you read http://www.unbound.net/documentation/howto_optimise.html ??
"Due to malloc overhead, the total memory usage is likely to rise to double
(or 2.5x) the total cache memory that is entered into the config."
And no, memory leaks are not common in Unbound to my knowledge at least.

Yes,I know them.
I have read the document.
But,because I have felt using too many memory by Unbound Process...

I think trying the tuning of parameters of value.
thanks.

best regards.

/// Maru

Post by l***@kwsoft.de

Hello,
have you read http://www.unbound.net/documentation/howto_optimise.html ??
"Due to malloc overhead, the total memory usage is likely to rise to double
(or 2.5x) the total cache memory that is entered into the config."
And no, memory leaks are not common in Unbound to my knowledge at least.
Regards
Andreas
_______________________________________________
Unbound-users mailing list
http://unbound.nlnetlabs.nl/mailman/listinfo/unbound-users

Rygl Aleš

2015-05-24 05:20:37 UTC

Permalink

Hi,

we are running 1.4.22 for more than a year. Unbound is allowed to use ~70 GiB
(and uses) of memory and we never had a problem with memory leaks. The same
setup is on eight servers.

Ales

Techs_Maru

2015-05-26 02:52:33 UTC

Permalink

Hi,

Thank you for teaching me.

Post by Rygl AleÅ¡
we are running 1.4.22 for more than a year. Unbound is allowed to use ~70 GiB
(and uses) of memory and we never had a problem with memory leaks. The same
setup is on eight servers.

Please tell me if that is possible,
How about value is rrset-cache-size? and prefetch parameter is on?

Best regards.

/// Maru

Post by Rygl AleÅ¡
Hi,
we are running 1.4.22 for more than a year. Unbound is allowed to use ~70 GiB
(and uses) of memory and we never had a problem with memory leaks. The same
setup is on eight servers.
Ales

Aleš Rygl

2015-05-26 06:51:54 UTC

Permalink

Hi,

see the common part of the config of my resolvers:

server:

verbosity: 1
statistics-interval: 900
statistics-cumulative: yes

extended-statistics: yes
num-threads: 16
msg-cache-size: 18G

msg-cache-slabs: 16
num-queries-per-thread: 2048
rrset-cache-size:
36G
rrset-cache-slabs: 16
infra-cache-slabs: 16
infra-cache-numhosts:
1000000
do-ip6: yes
access-control: 0.0.0.0/0 allow
access-control:
::0/0 allow
logfile: "/var/log/unbound/unbound.log"
log-time-ascii:
yes
hide-identity: yes
hide-version: yes
version: ""

use-caps-for-id: yes
prefetch: yes
dlv-anchor-file: "dlv.isc.org.key"

val-clean-additional: yes
val-log-level: 2
key-cache-size: 12G

key-cache-slabs: 16
neg-cache-size: 512m
local-data:
"z0vWaTC0.testdomain.test. IN A 10.0.0.1"
include:
"/etc/unbound/unbound.conf.d/local_zones"
include:
"/etc/unbound/unbound.conf.d/forward_zones"

python:
remote-control:

control-enable: yes
control-interface: 127.0.0.1
control-port: 953

On

Hi,
Thank

you for teaching me.

Post by Rygl AleÅ¡
we are running 1.4.22 for more than a year.

Unbound is allowed to use ~70 GiB (and uses) of memory and we never had
a problem with memory leaks. The same setup is on eight servers.
Please tell me if that is possible, How about value is rrset-cache-size?
and prefetch parameter is on?

Eduardo Schoedler

2015-05-26 18:35:20 UTC

Permalink

WOW, can you describe why you need such a large setup? (70GB seems a
bit of overkill for 1 node).

--
Eduardo Schoedler

Post by AleÅ¡ Rygl
Hi,
verbosity: 1
statistics-interval: 900
statistics-cumulative: yes
extended-statistics: yes
num-threads: 16
msg-cache-size: 18G
msg-cache-slabs: 16
num-queries-per-thread: 2048
rrset-cache-size: 36G
rrset-cache-slabs: 16
infra-cache-slabs: 16
infra-cache-numhosts: 1000000
do-ip6: yes
access-control: 0.0.0.0/0 allow
access-control: ::0/0 allow
logfile: "/var/log/unbound/unbound.log"
log-time-ascii: yes
hide-identity: yes
hide-version: yes
version: ""
use-caps-for-id: yes
prefetch: yes
dlv-anchor-file: "dlv.isc.org.key"
val-clean-additional: yes
val-log-level: 2
key-cache-size: 12G
key-cache-slabs: 16
neg-cache-size: 512m
local-data: "z0vWaTC0.testdomain.test. IN A 10.0.0.1"
include: "/etc/unbound/unbound.conf.d/local_zones"
include: "/etc/unbound/unbound.conf.d/forward_zones"
control-enable: yes
control-interface: 127.0.0.1
control-port: 953
Hi,
Thank you for teaching me.
we are running 1.4.22 for more than a year. Unbound is allowed to use ~70 GiB
(and uses) of memory and we never had a problem with memory leaks. The same
setup is on eight servers.
Please tell me if that is possible,
How about value is rrset-cache-size? and prefetch parameter is on?

Aleš Rygl

2015-05-27 14:29:47 UTC

Permalink

Well, we could live with much less, I agree. On the other hand
there is ~ 1mil. of active clients doing tens of thousands req/s. And we
have an excellent cache hit ratio, more than 90%.

On Tue, 26 May 2015

Post by Eduardo Schoedler
WOW, can you describe why

you need such a large setup? (70GB seems a

Post by Eduardo Schoedler
bit of overkill for 1

node).

Techs_Maru

2015-05-28 05:16:12 UTC

Permalink

Hi,

Thank you for showing config data.
It was very useful.

/// Maru