Thanks to Dave, Gene, Ana, Anthony for answering my query.
Jerson
--- Gene Chuang <[EMAIL PROTECTED]> wrote:
> Yup, that explains it! Weblogic admits to the diamond bug and has fixed it
> in sp5.
>
> Gene
>
> -----Original Message-----
> From: A mailing list for Enterprise JavaBeans development
> [mailto:[EMAIL PROTECTED]]On Behalf Of Jerson Chua
> Sent: Friday, September 01, 2000 3:20 AM
> To: [EMAIL PROTECTED]
> Subject: Re: Transaction
>
>
> Weblogic 5.1 with sp4. Im going to try sp5 on monday.
> --- Gene Chuang <[EMAIL PROTECTED]> wrote:
> > Hi,
> >
> > Which ejb server and what version are you using?
> >
> > Gene
> >
> > -----Original Message-----
> > From: A mailing list for Enterprise JavaBeans development
> > To: [EMAIL PROTECTED]
> > Sent: 8/31/00 11:48 PM
> > Subject: Re: Transaction
> >
> > Have anyone encountered similar problem? Is it normal that ejbstore is
> > deferred until the
> > transaction ends? My impression is that transaction isolation is
> > handled by the DB but
> > it seems to me that the application server is doing it instead.
> > Experts... please
> > comment on this.
> >
> > Jerson
> >
> > --- Rajan Kashyap <[EMAIL PROTECTED]> wrote:
> > > Hi Dave,
> > > I didn't get the exact meaning of
> > > "there can be two physcial instances of a single EJBObject of
> > "id" which is
> > > causing the local diamond to form."
> > >
> > > Can you please elaborate a bit more..
> > >
> > > Thanks,
> > > Rajan
> > >
> > > -----Original Message-----
> > > From: A mailing list for Enterprise JavaBeans development
> > > [mailto:[EMAIL PROTECTED]]On Behalf Of Dave Wolf
> > > Sent: Thursday, August 31, 2000 6:05 PM
> > > To: [EMAIL PROTECTED]
> > > Subject: Re: Transaction
> > >
> > >
> > > It appears you have a somewhat off topic version of a local diamond
> > forming.
> > > Here is the issue.
> > >
> > > methodA() {
> > > eb1 = ebhome.findByPK(id);
> > > // container calles ejbLoad() loading 111
> > > eb1.setXXX(222);
> > > // set is called on eb1 but since we are in tx
> > > // the container fails to call ejbStore() waiting for
> > a
> > > // tx boundry. Performance enhancement.....
> > > System.out.println(eb1.getXXX()); // returns 222
> > > // reading 222 from instance level data.
> > >
> > > // just tried this after my original post
> > > eb2 = ebhome.findByPK(id);
> > > // container calls ejbLoad() loading 111 as this is
> > what is
> > > // in the database as eb1 never had ejbStore() called
> > on it
> > > System.out.println(eb2.getXXX()); // returns 111 -
> > PROBLEM
> > > // reading 111 from instance level data
> > >
> > >
> > > return methodB(id);
> > >
> > > Remember there can be two physcial instances of a single EJBObject of
> > "id"
> > > which is causing the local diamond to form. The container should be
> > keeping
> > > track of every instance of an EJBOBject for id and assuring no diamond
> > > forms. In this case, the container should know a setXXX method was
> > called
> > > and ejbStore() is yet to be called, so when findByPrimaryKey() causes
> > the
> > > ejbLoad(), the container should force eb1 to call ejbStore() first to
> > > maintain consistancy.
> > >
> > > The diamond is now compounded because eb1 and eb2 have differing
> > states for
> > > XXX. If the container calls store() on eb1 then eb2 the value will
> > rever
> > > back to 111. Seems like 1) The container should have detected this
> > diamond
> > > and 2) its a good reason for optimistic locking :)
> > >
> > > Just my read on whats happening. Any other ideas anyone? Ive not
> > seen a
> > > diamond like this one before, so Im not 100% sure of my diagnosis but
> > it
> > > makes sense. The part I dont understand is why there was no locking
> > issue.
> > > That could only happen with eb1 and eb2 used the same connection?
> > That
> > > would be kinda odd too......
> > >
> > > Dave Wolf
> > > Internet Applications Division
> > > Sybase
> > >
> > >
> > >
> > >
> > > > -----Original Message-----
> > > > From: A mailing list for Enterprise JavaBeans development
> > > > [mailto:[EMAIL PROTECTED]]On Behalf Of Jerson Chua
> > > > Sent: Thursday, August 31, 2000 10:12 AM
> > > > To: [EMAIL PROTECTED]
> > > > Subject: Re: Transaction
> > > >
> > > >
> > > > Hi Dave...
> > > > The Session bean is stateless. The Entity bean is bean managed.
> > > > The entity bean is
> > > > properly implemented (ejbload and ejbstore).
> > > > All methods' transaction attribute are set to requires (this is
> > > > tentative), Isolation
> > > > level is set to serializable.
> > > >
> > > > here's a pseudo code
> > > >
> > > > // eb.xxx's original value is 111
> > > >
> > > > class StatelessSession {
> > > >
> > > > methodA() {
> > > > eb1 = ebhome.findByPK(id);
> > > > eb1.setXXX(222);
> > > > System.out.println(eb1.getXXX()); // returns 222
> > > >
> > > > // just tried this after my original post
> > > > eb2 = ebhome.findByPK(id);
> > > > System.out.println(eb2.getXXX()); // returns 111 -
> > PROBLEM
> > > >
> > > > return methodB(id);
> > > > }
> > > >
> > > > methodB(id) {
> > > > eb3 = ebhome.findByPK(id);
> > > > System.out.println(eb3.getXXX()); // returns 111 -
> > PROBLEM
> > > > return eb3.getYYY(); // return a value object
> > > > which contains subset of the entity
> > > > bean's atrributes
> > > > }
> > > > }
> > > >
> > > > Is it because the isolation level is serializable? I think the
> > > > change made in methodA
> > > > should reflect in methodb since both methods are under the same
> > > > transactional context.
> > > >
> > > > Am I missing something? please guide me.
> > > >
> > > > thanks in advanced.
> > > > Jerson
> > > >
> > > >
> > > > --- Dave Wolf <[EMAIL PROTECTED]> wrote:
> > > > > Can I get a little better description? Is this BMP or CMP? Is
> > > > methodA()
> > > > > calling methodB() or is the client calling methodB() right
> > > > after methodA().
> > > > > What Tx level are both beans set to? I assume we have a
> > > > session bean which
> > > > > contains methodA() and methodB()? Stateful or stateless?
> > > > >
> > > > > Dave Wolf
> > > > > Internet Applications Division
> > > > > Sybase
> > > > >
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: A mailing list for Enterprise JavaBeans development
> > > > > > [mailto:[EMAIL PROTECTED]]On Behalf Of Jerson Chua
> > > > > > Sent: Thursday, August 31, 2000 5:45 AM
> > > > > > To: [EMAIL PROTECTED]
> > > > > > Subject: Transaction
> > > > > >
> > > > > >
> > > > > > Hi...
>
=== message truncated ===
__________________________________________________
Do You Yahoo!?
Yahoo! Mail - Free email you can access from anywhere!
http://mail.yahoo.com/
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff EJB-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
