Re: [Springframework-developer] Potentialt dangerous hibernate Session reuse

[Per O. <po...@no...>]

Hi Juergen,

Thank you for your comments.

On Monday 14 February 2005 16:13, Juergen Hoeller wrote:
>
> I am aware of the Hibernate docs that the Session needs to be discarded
> when an exception was thrown. However, I find that an overly strict
> statement (just like "there is no such thing as a non-transactional
> Session", their famous statement from their forums mid last year).

Okay, i tend to agree about it being an overly strict statement, also because 
I've looked into SessionImpl.clear() and it does what we want :-) ... but: 
When they've explicitly stated that a Session should not be reused in case of 
exception, upcoming implementations of Session might break spring then. Can 
it not?

> Session.clear should be good enough for resetting an OSIV Session after a
> rollback. It's probably advisable to not use OSIV in single session mode at
> all if you want to avoid side effects completely. (Closing the Session and
> opening a new one during OSIV doesn't really add value here.)

It will add the value that:
  1) the failing session is thrown away
  2) new calls into spring-managed beans will operate fine on the new session

I know, that lazy-loading of objects from dead session will fail.

> Have you considered using OpenSessionInViewFilter in "deferred close mode",
> that is, with the "singleSession" flag turned off? In that case, each
> transaction will use its own Hibernate Session, but all of those Sessions
> will be kept open until view rendering has completed.

Yes, actually I have tried it out. But this solution does not scale very well. 
A JDBC connection is assigned to each hibernate Session, so using deferred 
close uses a connection for each session. Hence, one web request will quickly 
use many connections, at the same time.

Per

-- 
Per Olesen @ Nordija A/S - www.nordija.com - main#: +45 70 20 25 10
email: po...@no... - cell#: +45 23 38 95 81