No, but pretty much anything read off the filesystem is eligible for the
cache (unless you're using raw devices).

If you want to reduce the cache and your application is reasonably well
behaved AND you have limits on the WebSphere heap you should be able to
reduce your virtual machine size.  Maybe drop it by 48M or 64M.

On Wed, 2004-07-28 at 10:21, James Melin wrote:
> Ahhh. Well we're not doing anything like that. I was watching the cache and
> it fell before any swapping happened. Still overing at 99.6 megs from a
> peak of 114. Not swapping to disk at all yet. I'm just wondering what it is
> caching since all we have are JSP's. java related stuff and application
> image files etc that are being held locally.  Is there anyway to
> interrogate the cache?
>
>
>
>
>              Rich Smrcina
>              <[EMAIL PROTECTED]
>              o.com>                                                     To
>              Sent by: Linux on         [EMAIL PROTECTED]
>              390 Port                                                   cc
>              <[EMAIL PROTECTED]
>              IST.EDU>                                              Subject
>                                        Re: Under what conditions will
>                                        Linux start releasing Cache?
>              07/28/2004 10:08
>              AM
>
>
>              Please respond to
>              Linux on 390 Port
>              <[EMAIL PROTECTED]
>                  IST.EDU>
>
>
>
>
>
>
> The cache will grow as the data blocks are read in from the filesystem.
> That is all the cache is used for, storing filesystem data in an easily
> accessed place.  When an application needs more memory the cache will be
> released to a point.
>
> The symptom that we were seeing was the paging device filling up and
> cache being depleted.  The virtual machine would usually go into a
> loop.  Internally, WebSphere was doing some thrashing on it's own.  It
> was basically in continuous garbage collection, trying to release
> storage for the needs of the application.
>
> It turned out that the application was doing a DB2 query and under
> certain circumstances did not put any conditions on the query, so the
> entire table would be read into an object (a 120MB object!).  WebSphere
> dutifully tried to acquire enough storage to hold the whole table, but
> was not able to do so.  It would start garbage collection, release a few
> objects, then some more rows could be retrieved, on and on, until it
> couldn't release any more storage.
>
> On Wed, 2004-07-28 at 09:51, James Melin wrote:
> > This is Websphere.  What would cause cache to grow constantly, from an
> > application perspective. What did you  see in your application?
> >
> >
> >
> >              Rich Smrcina
> >              <[EMAIL PROTECTED]
> >              o.com>
> To
> >              Sent by: Linux on         [EMAIL PROTECTED]
> >              390 Port
> cc
> >              <[EMAIL PROTECTED]
> >              IST.EDU>
> Subject
> >                                        Re: Under what conditions will
> >                                        Linux start releasing Cache?
> >              07/28/2004 09:39
> >              AM
> >
> >
> >              Please respond to
> >              Linux on 390 Port
> >              <[EMAIL PROTECTED]
> >                  IST.EDU>
> >
> >
> >
> >
> >
> >
> > What is this machine used for?  110 MB of cache is pretty much a waste
> > of storage.  If the machine eats into all (or most) of that before it
> > finally croaks, then you have a memory hog or leak somewhere.
> >
> > I've seen that happen with WebSphere and it turned out to be an
> > application problem.
> >
> > The cache will slowly get released to a certain point, then paging will
> > begin and the paging/cacheing dance will begin.  When the page device
> > fills and the cache is gone then your up the creek...
> >
> > On Wed, 2004-07-28 at 09:25, James Melin wrote:
> > > I've got a linux guest that is pushing the wall, 17 megs free and
> > dropping.
> > > Swapping is going to be inevitable. The problem is that the memory
> usage
> > > growth is slow, and that cache grows to 110 meg before we start to
> > approach
> > > the wall. Will SLES8 start to release or clean the cache before
> swapping
> > > becomes rampant?
> > >
> > > <vent_spleen>
> > >
> > > Since there seems to be no way to control the cache in SLES8, which
> > really
> > > sucks since, in our situation, it is caching unimportant things. None
> of
> > > the application data is local. If there are any of the Linux Kernal
> > > developers monitoring this list, add my voice to those that are telling
> > you
> > > that not being able to control cache behavior is a  really frustrating
> > when
> > > working in a virtualized environment where you are trying to divide
> > limited
> > > resources amongst many guests. The Cache behaviour is the biggest
> > > non-application pig that I can see.
> > >
> > > Just my 2 cents
> > >
> > > </vent_spleen>
> > >
> > > ----------------------------------------------------------------------
> > > For LINUX-390 subscribe / signoff / archive access instructions,
> > > send email to [EMAIL PROTECTED] with the message: INFO LINUX-390
> or
> > visit
> > > http://www.marist.edu/htbin/wlvindex?LINUX-390
> > --
> > Rich Smrcina
> > illustro Systems International, LLC
> >                     --- See The Light---
> > Visit www.illustro.com to experience:
> >     z/Web-Host -- Easy Web-enabling for your Mainframe
> >     z/XML-Host -- Easy XML Enablement for your Mainframe
> >
> > Tel: +1.214.800.8900  Fax: +1.214.800.8989
> >
> > Catch the WAVV!  http://www.wavv.org
> > WAVV 2005 - Colorado Springs - May 20-24, 2005
> >
> > ----------------------------------------------------------------------
> > For LINUX-390 subscribe / signoff / archive access instructions,
> > send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
> > visit
> > http://www.marist.edu/htbin/wlvindex?LINUX-390
> >
> > ----------------------------------------------------------------------
> > For LINUX-390 subscribe / signoff / archive access instructions,
> > send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
> visit
> > http://www.marist.edu/htbin/wlvindex?LINUX-390
> --
> Rich Smrcina
> illustro Systems International, LLC
>                     --- See The Light---
> Visit www.illustro.com to experience:
>     z/Web-Host -- Easy Web-enabling for your Mainframe
>     z/XML-Host -- Easy XML Enablement for your Mainframe
>
> Tel: +1.214.800.8900  Fax: +1.214.800.8989
>
> Catch the WAVV!  http://www.wavv.org
> WAVV 2005 - Colorado Springs - May 20-24, 2005
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or
> visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
>
> ----------------------------------------------------------------------
> For LINUX-390 subscribe / signoff / archive access instructions,
> send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
> http://www.marist.edu/htbin/wlvindex?LINUX-390
--
Rich Smrcina
illustro Systems International, LLC
                    --- See The Light---
Visit www.illustro.com to experience:
    z/Web-Host -- Easy Web-enabling for your Mainframe
    z/XML-Host -- Easy XML Enablement for your Mainframe

Tel: +1.214.800.8900  Fax: +1.214.800.8989

Catch the WAVV!  http://www.wavv.org
WAVV 2005 - Colorado Springs - May 20-24, 2005

----------------------------------------------------------------------
For LINUX-390 subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit
http://www.marist.edu/htbin/wlvindex?LINUX-390

Reply via email to