[hibernate-dev] hibernate-osgi JPA bootstrap & classloader

Steve Ebersole steve at hibernate.org
Wed May 27 23:43:41 EDT 2015


On Wed, May 27, 2015 at 9:09 PM, Brett Meyer <brmeyer at redhat.com> wrote:

> > Is that the best way to know when TransactionManagers and DataSources
> come
> > and go too?  Or is there a more specific concept for listening to an
> "OSGi
> > service"?
>
> At least for TransactionManagers, yes, the BundleListener is probably the
> best approach.  I'm not aware of a more specific way beyond that, but I'm
> not 100%.  Honestly not sure about the DataSource -- can't remember if most
> containers automatically expose that or not.
>

But I would have to know the bundle that TransactionManager comes from in
order to listener to it, no?

I think we ultimately want an "OSGi DataSource adpater" as well, sort of
like we do with OsgiJtaPlatform.  I say an adapter here so that it can fit
with both ConnectionProvider and with MultiTenantConnectionProvider.


> >
> > Also, do the containers generally handle "in-flight" requests in
> > regards to TransactionManagers
> > and DataSources on activated/deactivated?  I guess this is more a
> > curiosity, since I do not see what we could do if the container does not
> > manage it.
>
> Really good question, but not sure.  The hope is that a truly
> OSGi-friendly bundle is "dynamic" enough to gracefully terminate
> in-progress actions if deactivated.  They're given that opportunity through
> BundleActivator#stop.  Of course, that's entirely up to the specific impl.
> I would expect Karaf container-provided defaults (Geronimo JTA, etc.) to
> handle that reasonably well for basic cases, but not sure about the
> real-world edge cases...
>
> Note that I always hoped to make Hibernate fit that bill, but we're *far*
> from it.  Deactivation would, in all likelihood, be catastrophic.  And no
> idea about re-activation.
>

I *think* de-activation would be catastrophic during the period that the
depended-on bundle is not available.  I am ok with that.  That seems
natural to me.  If a bundle we are providing persistence for says it wants
to use a JTA TransactionManager e.g. and that TransactionManager is not
available... I think an exception is perfectly valid outcome.

But yes it is the re-activation case (the TransactionManager becomes
available again) that I want to work.  In theory this should be a simple
thing to do; I think we already have the proper encapsulation with
OsgiJtaPlatform.  I just have no idea how to know when these things become
available and when they become unavailable.

I'll poke around on the Kara user list and see what I can find.


More information about the hibernate-dev mailing list