Hi there,
now I'm notice 1.7.3 eats too much RSS memory after ~1 week of non-stop
unning.
It seems 523 Mb too much instead ~200 Mb of previous version with same
config.
Is there memory leak occurs?
WBR, Yuri
Hi there,
now I'm notice 1.7.3 eats too much RSS memory after ~1 week of non-stop
unning.It seems 523 Mb too much instead ~200 Mb of previous version with same
config.Is there memory leak occurs?
WBR, Yuri
It has been driving me bonkers as I cannot pin it down but I am
experiencing an ever increasing memory usage on a kernel 4.17.4 system
and suspecting unbound 1.7.3 being the culprit.
For some reason a similar setup on a kernel 4.4.138 does no exhibit such
symptom.
09.07.2018 02:15, ѽ҉ᶬḳ℠ via Unbound-users пишет:
Hi there,
now I'm notice 1.7.3 eats too much RSS memory after ~1 week of non-stop
unning.It seems 523 Mb too much instead ~200 Mb of previous version with same
config.Is there memory leak occurs?
WBR, Yuri
It has been driving me bonkers as I cannot pin it down but I am
experiencing an ever increasing memory usage on a kernel 4.17.4 system
and suspecting unbound 1.7.3 being the culprit.
For some reason a similar setup on a kernel 4.4.138 does no exhibit such
symptom.
I'm using Solaris 10, not any Linux. So, this is not solution.
It has been driving me bonkers as I cannot pin it down but I am
experiencing an ever increasing memory usage on a kernel 4.17.4 system
and suspecting unbound 1.7.3 being the culprit.
For some reason a similar setup on a kernel 4.4.138 does no exhibit such
symptom.I'm using Solaris 10, not any Linux. So, this is not solution.
Perhaps a misunderstanding due to my phrasing. What I meant is that one
my systems is exhibiting some memory leakage too whilst the other is not.
Not sure this is OS issue. This time can't do comparison with other
servers with different platform.
Just guessing leakage.
09.07.2018 02:22, ѽ҉ᶬḳ℠ via Unbound-users пишет:
Not sure this is OS issue. This time can't do comparison with other
servers with different platform.Just guessing leakage.
From what I am observing I would concur. The difference though seems
that this end the amount of cached (~ 110 MB) / rss (~ 2MB) memory
remains stable/stagnant but mem_usage keeps on increasing
steadily/relentlessly to no end, well the end being out of memory and in
order to prevent that had to restart unbound already.
Except for dhcp client / dns resolver / time sync there nothing else
running on the box other than unbound 1.7.3
Hi,
Not sure this is OS issue. This time can't do comparison with other
servers with different platform.Just guessing leakage.
From what I am observing I would concur. The difference though seems
that this end the amount of cached (~ 110 MB) / rss (~ 2MB) memory
remains stable/stagnant but mem_usage keeps on increasing
steadily/relentlessly to no end, well the end being out of memory and in
order to prevent that had to restart unbound already.
The memory configuration is for the amount that Unbound allocates. The
total max amount of memory that Unbound ends up using because of that is
about double the specified amount (this may differ for different
implementations of malloc on different systems). So if you specified
100 Mb cache, it'll end up using 200Mb of memory for that. I have
reports that it then stabilizes and stops growing, for others. Add up
the memory amounts for the caches and x2 is the guess I would make. In
reality unbound also needs some memory for system buffers and so on, for
large TCP (or TLS) deployments that is a considerable amount. So perhaps
it is that you configured a lot of memory to be used and it is simply
growing to fill the config amount?
Best regards, Wouter
The memory configuration is for the amount that Unbound allocates. The
total max amount of memory that Unbound ends up using because of that is
about double the specified amount (this may differ for different
implementations of malloc on different systems). So if you specified
100 Mb cache, it'll end up using 200Mb of memory for that. I have
reports that it then stabilizes and stops growing, for others. Add up
the memory amounts for the caches and x2 is the guess I would make. In
reality unbound also needs some memory for system buffers and so on, for
large TCP (or TLS) deployments that is a considerable amount. So perhaps
it is that you configured a lot of memory to be used and it is simply
growing to fill the config amount?Best regards, Wouter
Thank you for the feedback. Below is the deployed configuration. Over
the past 12 hours the ram available reported by >>free -m<< decreased
from 1885 to 1854, that is ~ 2.5 MB/hrs
rrset-roundrobin: yes
outgoing-num-tcp: 25
incoming-num-tcp: 25
do-ip6: no
so-reuseport: yes
so-rcvbuf: 4m
so-sndbuf: 4m
use-caps-for-id: yes
qname-minimisation: yes
harden-below-nxdomain: yes
key-cache-size: 50m
key-cache-slabs: 1
neg-cache-size: 200k
msg-cache-slabs: 1
num-queries-per-thread: 50
rrset-cache-size: 100m
rrset-cache-slabs: 1
infra-cache-slabs: 1
infra-cache-numhosts: 200
msg-buffer-size: 8192