SWAP

[Davi Loperz <at...@gmail.com>]

I'm running ATS upon a dual xeon server with 24G of RAM.
Traffic approximately 50Mb.
With less than 30 minutes for the server to use all the memory, going to
swap memory, impairing the functioning of the machine.
Have you reviewed the file records.config but found nothing that could
cause the problem.
Have not even configured to use ram and still the problem persists.
What is causing this inconvenience can?


Thanks.
Virtual Word Network Company

Re: SWAP

[Davi Loperz <at...@gmail.com>]

Memory compression is disable.


2012/12/14 Luca Rea <lu...@contactlab.com>

> If you have enabled memory compression can you try disabling it?
>



-- 

Thanks in advance.
Virtual Word Network Company

RE: SWAP

[Luca Rea <lu...@contactlab.com>]

If you have enabled memory compression can you try disabling it?

Re: SWAP

[Davi Loperz <at...@gmail.com>]

There are 5 discs in storage.config.
3 discs 1T and 2T two discs.


2012/12/14 Luca Rea <lu...@contactlab.com>

> What is your storage.config?
>



-- 

Thanks in advance.
Virtual Word Network Company

RE: SWAP

[Luca Rea <lu...@contactlab.com>]

What is your storage.config?

Re: SWAP

["ming.zym@gmail.com" <mi...@gmail.com>]

hmm, every one reporting the problems should at least include: ATS
version and OS informations, and content and traffic informations, that
will help us identify the problem quickly.

in this case I will assume that you are running ATS v3.2 and x86_64.
that is the mostly used system in production. correct me if you don't
use that way, or change it right now.

in your case, the default RAM cache will consume about 7G ram, that is
1/1000 * 7TB. and the index of storage, we take the default 8KB as the
average size, will need be at least 7G, 7TB/8KB*10, that is 14G memory
usage at the end. in your case, you should need to tweak:
CONFIG proxy.config.cache.min_average_object_size INT 8000
to larger size, according to your content average size, that will cut
down memory usage if you have big files in ATS, I think that is your
case, if you are using disks > 1TB :D


then if it is not the memory shortage in the design, you are have some
real memory problems, there are at least two directions:
1, ram_cache:
  by default, the ram cache algorithm is set to CLFUS, which will prefer
the small file other than big files, which will cause memory problems if
you do not leave some memory buffer for the whole system, we find that
change to the standard LRU will help in this case. more detail on
TS-1006
config change to: 
CONFIG proxy.config.cache.ram_cache.algorithm INT 1

2, memory leak in every request
 there is another memory leak issue. be sure to fill as much information
as possible. be sure to upgrade to the git master codes and to another
test, enable memory usage dumper in recodes.config:
CONFIG proxy.config.dump_mem_info_frequency INT 30
then collect the results from logs
open a bug in the official jira system


FYI

-------- 原始信息 --------
发件人: Davi Loperz <at...@gmail.com>
Reply-to: users@trafficserver.apache.org
收件人: users@trafficserver.apache.org
主题: SWAP
日期: Fri, 14 Dec 2012 07:52:15 -0300

I'm running ATS upon a dual xeon server with 24G of RAM.
Traffic approximately 50Mb.
With less than 30 minutes for the server to use all the memory, going to
swap memory, impairing the functioning of the machine.
Have you reviewed the file records.config but found nothing that could
cause the problem.
Have not even configured to use ram and still the problem persists.
What is causing this inconvenience can?



Thanks.
Virtual Word Network Company


-- 
Zhao Yongming
aka: zym @ apache.org, yonghao @ taobao.com