Plone

Vito wrote:
> 
> 
> Usually I've to reboot the serve cause the python processes burn all the
> swap without "touch" many RAM.
> 
> 

This is truly a nonsense statement. 

Please explain what you really mean and how you back this statement with
evidence.

-aj
-- 
View this message in context: http://plone.293351.n2.nabble.com/Plone-burns-swap-memory-but-it-uses-few-RAM-why-tp5962790p5962852.html
Sent from the General Questions mailing list archive at Nabble.com.

On Jan 26, 10:12 am, Vito <vitofa...@...> wrote:

> I've a server dual core with 4 GB di RAM. This server is used only for host
> 2 zope instances with a plone site for each (and pound + varnish, with
> apache as frontend)
>
> This two plone sites have a standard configuration with few editors (20).
> Usually I've to reboot the serve cause the python processes burn all the
> swap without "touch" many RAM.
>
> Why? Can i solve this problem without a reboot every night?

Andreas is being his usual abrasive self, but also as usual, he _is_
right.  Technically, it is not possible for your system to use swap
without using RAM.  Swappers move data _from_ RAM to disk, to make
room for other tasks which need the RAM.  If your swap is full, and
your physical memory is not being used, it simply means some process
with large memory requirements has started (causing other memory to be
swapped out) & stopped (thus freeing the physical memory it was
using), and nothing has needed to call the swapped data back into
memory since.

However, even if there is zero space left in your swap file, if you
aren't using 4GB of RAM at the same time, there is no need to restart
your server - everything is working as intended.  If your system has
_never_ used all of its physical memory, you have no problem.  You
only have an issue if you run into situations where all of your
physical memory is used, all of your swap is used, and increasing the
size of your swap file doesn't improve performance.

Derek Broughton-3 wrote:
> 
> On Jan 26, 10:12 am, Vito <vitofa...@...> wrote:
> 
>> I've a server dual core with 4 GB di RAM. This server is used only for
>> host
>> 2 zope instances with a plone site for each (and pound + varnish, with
>> apache as frontend)
>>
>> This two plone sites have a standard configuration with few editors (20).
>> Usually I've to reboot the serve cause the python processes burn all the
>> swap without "touch" many RAM.
>>
>> Why? Can i solve this problem without a reboot every night?
> 
> Andreas is being his usual abrasive self, but also as usual, he _is_
> right. 
> 

Well - if you are in the business of running and administrating a server
then one should know what swapping means and how memory management on a Unix
system is working in general. Such a wild claim from a person doing
adminstration is kind of scary.

-aj
-- 
View this message in context: http://plone.293351.n2.nabble.com/Plone-burns-swap-memory-but-it-uses-few-RAM-why-tp5962790p5963067.html
Sent from the General Questions mailing list archive at Nabble.com.

Vito wrote at 2011-1-26 06:12 -0800:
>I've a server dual core with 4 GB di RAM. This server is used only for host
>2 zope instances with a plone site for each (and pound + varnish, with
>apache as frontend)
>
>This two plone sites have a standard configuration with few editors (20).
>Usually I've to reboot the serve cause the python processes burn all the
>swap without "touch" many RAM.
>
>Why? Can i solve this problem without a reboot every night?

As often, I do not share Andreas opinions -- and find his remarks to
your post inadequate.

>From the distance, your observation is not easy to understand.

I can see 2 potential causes (but there are probably more):

  *  operating systems allow via configuration to control
     how memory is used. The default configuration (for Linux) is to
     use as many memory for processes as is available
     and start using swap only when there is a memory shortage.

     In such a configuration, one would expect that swap
     is filled only after almost all memory is used up.
     But, you might have a specially configured system?

  *  in some operating system configurations, "/tmp" and swap
     share a common space -- thus, when "/tmp" grows the available
     space for "swap" shrinks and vice versa.

     I have seen such configurations for Solaris -- but not yet
     for Linux.

Or, to say it differently: with a standard configuration, you
should not see what you have reported.


We do not have sufficient information to explain your observations
adequately.


--
Dieter

Dieter Maurer wrote at 2011-1-27 10:37 +0100:
>Vito wrote at 2011-1-26 06:12 -0800:
>>I've a server dual core with 4 GB di RAM. This server is used only for host
>>2 zope instances with a plone site for each (and pound + varnish, with
>>apache as frontend)
>>
>>This two plone sites have a standard configuration with few editors (20).
>>Usually I've to reboot the serve cause the python processes burn all the
>>swap without "touch" many RAM.
>>
>>Why? Can i solve this problem without a reboot every night?
> ...
>>From the distance, your observation is not easy to understand.
>
>I can see 2 potential causes (but there are probably more):
>
>  *  operating systems allow via configuration to control
>     how memory is used. The default configuration (for Linux) is to
>     use as many memory for processes as is available
>     and start using swap only when there is a memory shortage.
>
>     In such a configuration, one would expect that swap
>     is filled only after almost all memory is used up.
>     But, you might have a specially configured system?
>
>  *  in some operating system configurations, "/tmp" and swap
>     share a common space -- thus, when "/tmp" grows the available
>     space for "swap" shrinks and vice versa.
>
>     I have seen such configurations for Solaris -- but not yet
>     for Linux.

I just noticed an additional possibility:

   *  Some operating systems support so called RAM partitions:
      that is a partition that feels like a normal file system
      partition but actually is backed by RAM only (not the hard disk).

      Again, I know this from Solaris where this setup is occasionally
      used for "/tmp".

      In such a case, much less memory may be available for
      processes than one would expect because part is used
      by a RAM partition.

      Again, this is not a default configuration.


--
Dieter

Thank's all.

Unfortunately (? :) ) I'm not a sysadmin, but a "particular" customer with
an internal sysadmin staff told me the question I wrote in my above message.

I try to take a look considering your advices.

Vito

-- 
View this message in context: http://plone.293351.n2.nabble.com/Plone-burns-swap-memory-but-it-uses-few-RAM-why-tp5962790p5966296.html
Sent from the General Questions mailing list archive at Nabble.com.

Vito wrote:
> 
> Thank's all.
> 
> Unfortunately (? :) ) I'm not a sysadmin, but a "particular" customer with
> an internal sysadmin staff told me the question I wrote in my above
> message.
> 
> 

I would be scared having such a sysadmin staff.

-aj
-- 
View this message in context: http://plone.293351.n2.nabble.com/Plone-burns-swap-memory-but-it-uses-few-RAM-why-tp5962790p5966393.html
Sent from the General Questions mailing list archive at Nabble.com.

Vito wrote:
> 
> This two plone sites have a standard configuration with few editors (20).
> Usually I've to reboot the serve cause the python processes burn all the
> swap without "touch" many RAM.
> 
> Why? Can i solve this problem without a reboot every night?
> 

I wrote this many moons ago; I think it may help you:
http://hvelarde.blogspot.com/2007/12/zodb-performance-small-change-can-make.html

best regards
-- 
View this message in context: http://plone.293351.n2.nabble.com/Plone-burns-swap-memory-but-it-uses-few-RAM-why-tp5962790p5966853.html
Sent from the General Questions mailing list archive at Nabble.com.

Andreas,

On Thu, Jan 27, 2011 at 5:56 AM, ajung <lists@...> wrote:
> Vito wrote:
>
>> Unfortunately (? :) ) I'm not a sysadmin, but a "particular" customer with
>> an internal sysadmin staff told me the question I wrote in my above
>> message.
>
> I would be scared having such a sysadmin staff.

Please stop this line of discussion. Implicitly insulting the customer
and/or his sysadmin is not productive or helpful.

I'm sending this to the list instead of mailing you personally because
I want it to be clear to all that I don't think this kind of behavior
is appropriate in the Plone community, and I don't want others to
think that silence equals tolerance of this behavior.

:Jon

Jon Stahl <jonstahl@...> writes:

> Andreas,
>
> On Thu, Jan 27, 2011 at 5:56 AM, ajung <lists@...> wrote:
>> Vito wrote:
>>
>>> Unfortunately (? :) ) I'm not a sysadmin, but a "particular" customer with
>>> an internal sysadmin staff told me the question I wrote in my above
>>> message.
>>
>> I would be scared having such a sysadmin staff.
>
> Please stop this line of discussion. Implicitly insulting the customer
> and/or his sysadmin is not productive or helpful.
>
> I'm sending this to the list instead of mailing you personally because
> I want it to be clear to all that I don't think this kind of behavior
> is appropriate in the Plone community, and I don't want others to
> think that silence equals tolerance of this behavior.

+10 and thanks for writing so well and clearly on this.

Ross

On Jan 27, 5:37 am, "Dieter Maurer" <die...@...> wrote:

>   *  in some operating system configurations, "/tmp" and swap
>      share a common space -- thus, when "/tmp" grows the available
>      space for "swap" shrinks and vice versa.
>
>      I have seen such configurations for Solaris -- but not yet
>      for Linux.

Linux has had that for nearly a decade.  It doesn't alter the fact
that he doesn't have a problem unless he's completely out of virtual
memory.