I can tell you that the maxobjects discard algorithm is pretty.. well..
ech...

That may explain the CPU overuse.

I would start my bringing down the minutes, and see if that helps.

I've been working through a leak with Brian G, but have been unable to
reproduce it locally, so I have no idea what has happened there.  He seems
to have conquered it by using Java6u10.

Mark





On Sat, Sep 5, 2009 at 5:40 AM, Clint <[email protected]> wrote:

>
> Oh, I misunderstood you earlier. Yeah, we definitely had it off- all
> three app functions disabled, and the app itself was closed, so it
> wasn't connected to the runtime via Flex.
>
> Like I said, though, we experienced issues in both cases: Monitor on
> and Monitor off.
>
> On Sep 4, 2:33 pm, Bob Silverberg <[email protected]> wrote:
> > If you go to the Server Monitor, you'll see three buttons at the top
> > of the screen, one for each of:
> >
> > Monitoring, Profiling, Memory Tracking
> >
> > If the option is currently off, the button will read "Start xxx", if
> > it's currently on it will read "Stop xxx"
> >
> > Make sure that all of them are currently turned off.
> >
> > You can also do this via the admin api, using the stopMonitoring()
> > method.  More info is available here:
> http://livedocs.adobe.com/coldfusion/8/htmldocs/help.html?content=ser...
> >
> > Cheers,
> > Bob
> >
> >
> >
> >
> >
> > On Fri, Sep 4, 2009 at 3:15 PM, Clint<[email protected]> wrote:
> >
> > > How do you actually turn it off to completely disable it? We
> > > experienced issues both with Profiling and Memory Tracking on and off,
> > > but like you suggested, maybe the monitor was completely disabled?
> >
> > > On Sep 3, 10:50 pm, Bob Silverberg <[email protected]> wrote:
> > >> Were you perchance running the server monitor while these problems
> > >> were occurring?  I ran into a situation in which Transfer was having
> > >> major problems when I had the server monitor running in the
> > >> background, even though Profiling and Memory Tracking were turned off.
> > >>  I was led to believe prior to this that having just Monitoring on,
> > >> without Profiling and Memory Tracking, would not impact the server's
> > >> performance, but it did.  As soon as I turned off all monitoring my
> > >> problem went away.
> >
> > >> This may not apply to your situation, but if it does, try turning off
> > >> the monitor entirely and see what happens.
> >
> > >> Cheers,
> > >> Bob
> >
> > >> On Thu, Sep 3, 2009 at 11:28 PM, Clint<[email protected]> wrote:
> >
> > >> > I want to chime in and give some slightly tangential information
> about
> > >> > this issue we ran into. (I work with "Z" on this product):
> >
> > >> > Because we originally thought that our application code was leaking
> > >> > memory (and indeed, it may still be, and the memory profiling
> options
> > >> > on ColdFusion 7 (JVM 1.4 series) are limited, to say the least, we
> > >> > thought we would go ahead upgrade ColdFusion to the last version of
> > >> > 8.0.1 running on JVM 1.6. (This is all running on 32-bit Windows
> > >> > 2003).
> >
> > >> > Z made the necessary changes to our app and we upgraded our server
> to
> > >> > CF8. The app ran but not as fast as expected- we expected very fast
> > >> > performance based on many reports of heavy CFC-based apps. (We are
> > >> > aware of the classloader bug in JVM 1.6_04 and experienced the issue
> > >> > I'm describing on JVM 1.6_04, 1.6_11, and 1.6_16, and JVM 1.5_19).
> > >> > Even more surprising than the diminished application performance, it
> > >> > was never very long before the app server's CPU would be pegged at
> > >> > 100%. A few times this happened we were able to peek under the hood
> > >> > with Adobe's Server Monitor app. We saw at least one long running
> > >> > thread during these times referencing Transfer's
> SoftReferenceHandler.
> >
> > >> > Upon further investigation, we realized that after the app started
> up,
> > >> > under no load whatsoever the CPU usage would be 25%, 50%, 75%, or
> 100%
> > >> > (notable b/c this is a 4 core machine). As best as I can tell, the
> > >> > long running thread I observed was completely monopolizing the core
> on
> > >> > which it was running, preventing the application from serving any
> > >> > other requests.
> >
> > >> > After we rolled back to CF7 for current production, we provisioned a
> > >> > Windows 2003 VM on Amazon EC2 and installed CF8 and our application.
> > >> > We were not able to reproduce the problem that we described, and the
> > >> > performance was more or less as we expected on 8 (reasonably fast
> > >> > considering the minimal CPU power dedicated on the "smallest" EC2
> > >> > instance).
> >
> > >> > I don't know that any of this will help you, Mark, or anyone else
> > >> > here, but it is an interesting observation and I want to provide as
> > >> > much information as possible to aid in the resolution of this
> problem.
> >
> > >> > Thanks!
> >
> > >> > Clint
> >
> > >> >> > > Hi Everybody,
> >
> > >> >> > > We have an e-commerce website developed using Coldfusion 7,
> Model-
> > >> >> > > Glue, Coldspring and Transfer. Over the past months we have
> struggled
> > >> >> > > with what we thought was a "Memory Leak". We could see the jrun
> memory
> > >> >> > > usage climb until it just became unresponsive (queue a bunch of
> > >> >> > > requests and never serve them). We would then have to restart
> jrun and
> > >> >> > > repeat the process over and over again.
> >
> > >> >> > > After months of trying to figure out what was going on we
> finally
> > >> >> > > realized who the culprit was; the Transfer cache.
> >
> > >> >> > > Now, here's our original cache configuration:
> >
> > >> >> > > <objectCache>
> > >> >> > >        <defaultcache>
> > >> >> > >                <maxobjects value="60"/>
> > >> >> > >    <maxminutespersisted value="10" />
> > >> >> > >                <accessedminutestimeout value="10"/>
> > >> >> > >                <scope type="instance" />
> > >> >> > >        </defaultcache>
> >
> > >> >> > >        <!-- SOME CLASSES WE DONT WANT TO CACHE -->
> > >> >> > >        <cache class="Users.Orders">
> > >> >> > >                <scope type="none" />
> > >> >> > >        </cache>
> > >> >> > >        ....
> > >> >> > > </objectCache>
> >
> > >> >> > > It turns out leaving the cache on by default was killing us.
> Although
> > >> >> > > performance was always great, under heavy load our site would
> > >> >> > > sometimes have to go down every hour.
> >
> > >> >> > > After some experimenting we decided to turn the cache off by
> default
> > >> >> > > and be very conservative on the objects we cached. Our latest
> > >> >> > > transfer.xml looks something like this:
> >
> > >> >> > > <objectCache>
> > >> >> > >        <defaultcache>
> > >> >> > >                <scope type="none" />
> > >> >> > >        </defaultcache>
> > >> >> > >        <cache class="Users.Customer">
> > >> >> > >          <maxobjects value="60"/>
> > >> >> > >    <maxminutespersisted value="10" />
> > >> >> > >                <accessedminutestimeout value="10"/>
> > >> >> > >                <scope type="instance" />
> > >> >> > >        </cache>
> > >> >> > >        <cache class="Products.ProductType">
> > >> >> > >          <maxobjects value="10"/>
> > >> >> > >    <maxminutespersisted value="30" />
> > >> >> > >                <accessedminutestimeout value="10"/>
> > >> >> > >                <scope type="instance" />
> > >> >> > >        </cache>
> > >> >> > >        ...
> > >> >> > > </objectCache>
> >
> > >> >> > > I guess the main question I have is whether anyone here has had
> > >> >> > > similar experiences or has any rule's of thumb when it comes to
> > >> >> > > caching objects. We are still unsure on whether the behavior we
> were
> > >> >> > > experiencing was an actual memory leak or just poor cache
> management.
> > >> >> > > Any comments?
> >
> > >> >> > --
> > >> >> > E: [email protected]
> > >> >> > T:http://www.twitter.com/neurotic
> > >> >> > W:www.compoundtheory.com
> >
> > >> --
> > >> Bob Silverbergwww.silverwareconsulting.com
> >
> > --
> > Bob Silverbergwww.silverwareconsulting.com
> >
>


-- 
E: [email protected]
T: http://www.twitter.com/neurotic
W: www.compoundtheory.com

--~--~---------~--~----~------------~-------~--~----~
Before posting questions to the group please read:
http://groups.google.com/group/transfer-dev/web/how-to-ask-support-questions-on-transfer

You received this message because you are subscribed to the Google Groups 
"transfer-dev" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to 
[email protected]
For more options, visit this group at 
http://groups.google.com/group/transfer-dev?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to