Stephan - isn't this needed so JCR transactions can participate in a wide JTA transaction? I find the suggestion otherwise quite alarming !
On 2/21/07, Stefan Guggisberg <[EMAIL PROTECTED]> wrote:
On 2/21/07, KÖLL Claus <[EMAIL PROTECTED]> wrote: > hi stefan, > > thanks for your information .. > do you prefer to use the j2c adapter in a production environment > to get a xasession (atomic transaction) ? the problem for me is that we are running in a websphere environemnt > and i'm not able to run the j2c adapter on it @see JCR JCR-743. personally i don't think that it is a good idea to use managed connections underneath jackrabbit. the persistence layer of jackrabbit should imo be as close to the backend as possible, i.e. control the lifecycle of the jdbc connection used. but that's a different topic. > Maby somebody has experiences with websphere ?? i don't :) cheers stefan > > > -----Ursprüngliche Nachricht----- > Von: Stefan Guggisberg [mailto:[EMAIL PROTECTED] > Gesendet: Mittwoch, 21. Februar 2007 11:34 > An: [email protected] > Betreff: Re: Persistence Manager in Production Environment > > i'd suggest, for now, using jackrabbit's default configuration, i.e. > DerbyPersistenceManager, since it's been extensively tested and since > it is pretty fast (considerably faster than the oracle pm). > > cheers > stefan > > On 2/21/07, KÖLL Claus <[EMAIL PROTECTED]> wrote: > > We are on the road to use jackrabbit in our production environment but i'm not sure > > which persistence manager we should use. > > > > I have made some tests with the j2c adapter in combination with the oracle db pm > > to get a atomar transaction through the pm. > > A reason for db is also the easy way to backup/restore the data. > > > > On the other hand is the filesystem, no atomar transaction but it is fast. we think we get about 1-2 Million documents in jackrabbit > > and then the backup is no more really possible on filesystem beacuse if i use objectpersistenmanager i get about > > 6-10 files per node on fs (6*2Million files effectivity). On crash to recover these files take really long :-) > > > > A good message is that day will offer there pm's, so the node propertys will no more be stored in seperate files .. > > > > I hope i get some experience ... > > thanks > > claus > > >
