Mailing-List: contact dev-help@jackrabbit.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@jackrabbit.apache.org
Received-SPF: pass (asf.osuosl.org: local policy)
From: "Giota Karadimitriou" <Giota.Karadimitriou@eurodyn.com>
To: <dev@jackrabbit.apache.org>
Subject: RE: jackrabbit & clustering
Date: Wed, 3 May 2006 16:48:20 +0300
Message-ID: <005a01c66eb8$3d291c20$b6fdfe0a@gkaradimitr>
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: 7bit
In-Reply-To: <4458AE2B.1080306@gmx.net>
Importance: Normal

Thanks for your time and comments once more Marcel,

another idea would be that maybe I could send the item states 
from the shared item state manager of one node to the shared item state
managers of other nodes and validate their cache using the public
available methods "stateCreated/stateDiscarded/stateModified".
 
The question is whether this would suffice (sharing and revalidating the
state items among shareditemstatemanagers)or more things are needed to
ensure integrity. 

regards
Giota

> -----Original Message-----
> From: Marcel Reutegger [mailto:marcel.reutegger@gmx.net]
> Sent: Wednesday, May 03, 2006 4:21 PM
> To: dev@jackrabbit.apache.org
> Subject: Re: jackrabbit & clustering
> 
> Giota Karadimitriou wrote:
> > Thanks for the answers Marcel. One extra question: since shared item
> > state manager is entirely responsible for controlling integrity; is
> > there a way to check (by querying somehow the database) whether an
item
> > state is up to date and if not to just reload it from the database.
> > Couldn't this ideally work for cluster since everything in our
> > application is being saved in a common database? (Of course
performance
> > would go down with all these db calls but cache would become valid)
> 
> well, in theory you could retrieve the item state again from the
> persistence manager and compare the modification count with the item
> state that you already have. if it is the same you know its
up-to-date.
> but that may change at any given point in time. imo you need some sort
> of distributed locking to be sure that you only work on up-to-date
item
> states.
> 
> regards
>   marcel