 | | | cocoon-ehcache-1.data becomes very large | cocoon-ehcache-1.data becomes very large 2006-07-13 - By Ard Schrijvers
Back
>
>
> Ard Schrijvers wrote:
> >> On 7/13/06, Andreas Hartmann <andreas@(protected)> wrote:
> >>
> >>> ...<parameter name="overflow-to-disk" value="false"/>
> >>>
> >>> Is there a special reason why this is set to "false"?...
> >> It's because in that case we only want a short-term cache.
> There are more
> >> parameters which might help, they are read by the
> >>
> src/java/org/apache/cocoon/components/store/impl/EHDefaultStor
> e.java class
> >> IIUC.
> >
> > Beside the fact that I am (have to begin but is on my
> agenda) going to write
> > a best practice on these kind of things, I will try some to
> explain some
> > stuff here in short:
>
> Ard, thanks for your comprehensive answers!
> I'll take a closer look at your recommendations.
>
> Will you announce your how-to on the list?
Yes I will. I am though not yet sure in what kind of form I will write it (wiki
or document). I am also quite occupied but try to find time to work on it
Regards Ard
>
> -- Andreas
>
> >
> > First of all, make sure you have the EHDefaultStore.java
> from the trunk. This
> > one has some important patches.
> >
> > Now, a common configuration for your store would be:
> >
> > <store logger="core.store"> <parameter name="maxobjects"
> value="1000"/>
> > <parameter name="eternal" value="false"/> <parameter
> name="timeToLiveSeconds"
> > value="0"/> <!-- 1 day --> <parameter name="timeToIdleSeconds"
> > value="86400"/> <parameter name="overflow-to-disk"
> value="true"/> <parameter
> > name="diskpersistent" value="true"/> </store>
> >
> > This means, a maximum of 1000 object (keys+response) are
> kept in memory.
> > Since eternal is false, timeToLiveSeconds and
> timeToIdleSeconds are taken
> > into account. Is timeToLiveSeconds < timeToIdleSeconds, then only
> > timeToIdleSeconds matters. In the configuration above,
> after 1 day of not
> > being used, the cachekey is removed from cache.
> >
> > So, when overflow-to-disk is set to true, and you don't
> have eternal false
> > and timeToIdleSeconds > 0, then ones a cachekey is in
> memory/disk, it will
> > never be removed (apart from the StoreJanitor which when
> low on memory lets
> > the EHDefaultStore free its cache ---though this couldn't
> be implemented
> > worse because it just starts removing cachekeys/reponses
> from index 0 (which
> > are commonly the most important ones) instead of removing
> them according the
> > correct eviction policy. I have posted already to the
> ehcache mailinglist how
> > to find the LRU one, but not yet got an answer--- and I
> assume you dont use
> > eventAware cache actively removing keys )
> >
> > So, it is very well possible, to let your cache grow
> INFINITELY large.
> > Setting timeToIdleSeconds is important. Certainly when for
> example you have
> > cachekeys including current date (though, watch out with
> these kind of cache
> > practices because your cache could grow very large)
> >
> > diskpersistent is added to be configurable since short, but
> does not really
> > matter. It only indicates persistent cache between JVM
> restarts (used to be
> > default true).
> >
> > If you are working with a cocoon versions < 2.1.9, and you
> dont want to
> > switch, but want to use the latest EHDefaultStore, just add the
> > EHDefaultStore.java from trunk in your
> > src/java/country/yourcomapany/cocoon..., and add to user.roles:
> >
> > <role name="org.apache.excalibur.store.Store  " shorthand="store"
> >
> default-class="country.yourcompany.cocoon.components.store.imp
> l.EHDefaultStore"
> > />
> >
> > Now, just check it out configuring timeToIdleSeconds to 5
> secs or something.
> > Then look at the StatusGenerator, and you will see
> cachekeys being dumped
> > after 5 sec (by the way, dont refresh the StatusGenerator
> to often then,
> > because looking at the StatusGenerator uses the keys,
> implying they are not
> > idle for another 5 sec... :-) )
> >
> > Hope things are little clearer...
> >
> > Regards Ard
> >
> >> -Bertrand
> >>
> >>
> -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ----
> -- ---- To
> >> unsubscribe, e-mail: users-unsubscribe@(protected)
> For additional
> >> commands, e-mail: users-help@(protected)
> >>
> >>
>
>
> --
> Andreas Hartmann
> Wyona Inc. - Open Source Content Management - Apache Lenya
> http://www.wyona.com http://lenya.apache.org
> andreas.hartmann@(protected) andreas@(protected)
>
>
> -- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
> To unsubscribe, e-mail: users-unsubscribe@(protected)
> For additional commands, e-mail: users-help@(protected)
>
>
-- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ---- ------
To unsubscribe, e-mail: users-unsubscribe@(protected)
For additional commands, e-mail: users-help@(protected)
|
|
|