tomee-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Lars-Fredrik Smedberg <itsme...@gmail.com>
Subject Re: JDBC Transactions vs EJBs and @TransactionAttribute(...) and more...
Date Thu, 16 Oct 2014 13:38:56 GMT
Thanks for the answer

What exactly is involved in a checkDs1() call?

Regards
LF

On Thu, Oct 16, 2014 at 3:33 PM, Romain Manni-Bucau <
rmannibucau@tomitribe.com> wrote:

> 2014-10-16 15:01 GMT+02:00 Lars-Fredrik Smedberg <itsmeden@gmail.com>:
> > Thanks again, see below.
> >
> > On Thu, Oct 16, 2014 at 2:05 PM, Romain Manni-Bucau <
> > rmannibucau@tomitribe.com> wrote:
> >
> >> 2014-10-16 13:20 GMT+02:00 Lars-Fredrik Smedberg <itsmeden@gmail.com>:
> >> > Just one more thing I forgot to put in the other mail
> >> >
> >> > With "Depend, using ManagedConnection it will be green but it will not
> >> do what
> >> > you expect." you mean that I will think it runs in a tx but its not
> and
> >> in
> >> > case of failure I might get inconsistency?
> >> >
> >>
> >> right, if datasource2.commit() fails but ds1.commit() passed then you
> >> are not consistent. Of course with jpa it is not that impacting since
> >>
> >
> > Could you please explain how jpa tries to solve this? I guess it will not
> > completely remove the consisitency risk but only lower the risk?
> >
>
> It doesn't but it limits it in practise. It basically do a check
> before calling commit. All checks pass before commit:
>
> checkDs1();
> checkDs2();
> commitDs1();
> commitDs2();
>
> so it avoids half of the ds2 commit fails but not ds1.
>
> >
> >> the provider checks it before committing.
> >>
> >>
> > So I guess in our case: non xa datasources with multiple connections
> within
> > a transaction we are not guaranteed consistency?
> >
>
> Yep
>
> > I see also that some db vendors (MQ SQL) has some optimization so
> > transactions involving multiple connections from different datasources
> but
> > from the same db is not promoted to a distributed tx
> >
> > => Is that only vendor specific or is that according to spec to not
> promote
> > it in certain situations?
> >
>
> vendor
>
> >
> > resource = datasource != underlying db (without using openejb
> >> aliases). Each datasource has its own pool of connection and then its
> >> own JTA integration (different Xid).
> >>
> >> > Regards
> >> > LF
> >> >
> >> > On Thu, Oct 16, 2014 at 12:59 PM, Romain Manni-Bucau <
> >> > rmannibucau@tomitribe.com> wrote:
> >> >
> >> >> Well I don't know websphere enough to answer to the related question
> >> >>
> >> >>
> >> >> 2014-10-16 12:55 GMT+02:00 Lars-Fredrik Smedberg <itsmeden@gmail.com
> >:
> >> >> > Thanks Romain, please see if you can help me on the follow up
> below?
> >> >> >
> >> >> >
> >> >> > On Thu, Oct 16, 2014 at 9:33 AM, Romain Manni-Bucau <
> >> >> rmannibucau@gmail.com>
> >> >> > wrote:
> >> >> >
> >> >> >> 2014-10-16 9:31 GMT+02:00 Lars-Fredrik Smedberg <
> itsmeden@gmail.com
> >> >:
> >> >> >> > @Romain
> >> >> >> >
> >> >> >> > I can see in the ManagedConnection code that the reused
> Connection
> >> is
> >> >> per
> >> >> >> > Transaction per DataSource.
> >> >> >> >
> >> >> >> > So if I use 2 x DataSource (pointing to the same DB but
with
> >> different
> >> >> >> > properties set) I will within a transaction use 2 connections,
> >> >> correct?
> >> >> >> >
> >> >> >> yep
> >> >> >>
> >> >> >> > Will this result in a distributed transaction?
> >> >> >> >
> >> >> >>
> >> >> >> No, this ManagedConnection is to get JTA integration "locally"
> (ie we
> >> >> >> integrate locally with XA hooks but it is not distributed).
For
> that
> >> >> >> you have to use a really XA datasource.
> >> >> >>
> >> >> >
> >> >> > We use WebSphere in test/production and the driver included in
> >> >> db2jcc4.jar.
> >> >> >
> >> >> > As far as I understand from the documentation if the type
> attribute of
> >> >> the
> >> >> > datasource element in the configuration is not set it will search
> the
> >> >> > driver jar for datasources in the following order:
> >> >> >
> >> >> > javax.sql.ConnectionPoolDataSource
> >> >> > javax.sql.DataSource
> >> >> > javax.sql.XADataSource
> >> >> >
> >> >> > In our testconfiguration we do not specify the type attribute
so I
> >> assume
> >> >> > we are NOT using XA datasource. correct?
> >> >> >
> >> >> > So some questions on the above:
> >> >> >
> >> >> > 1. In what cases will I not need a distributed tx?
> >> >>
> >> >> You don't need it either when you have a single or totally
> decorelated
> >> >> resources (datasource for instane) or when you want performances (XA
> >> >> is slow) and can accept some manual operations to fix inconsistency
> if
> >> >> it happens (should be rare)
> >> >>
> >> >> > a) A tx with 1 connection to a DB? No!
> >> >> > b) A tx with 2 connections (from seperate datasources) connected
to
> >> the
> >> >> > same DB? <== ??
> >> >> > c) A tx with 2 connections (from seperate datasources) connected
to
> >> >> > different DBs? Yes!
> >> >> > d) A tx with 1 connection to a DB and e.g. JMS? Yes!
> >> >> >
> >> >> > 2. If I have a scenario that requires a distributed and the
> datasource
> >> >> that
> >> >> > is used is NOT an XA datasource what will happen? Exception?
> >> >> >
> >> >>
> >> >> Depend, using ManagedConnection it will be green but it will not do
> >> >> what you expect.
> >> >>
> >> >> > Thanks
> >> >> >
> >> >> > Regards
> >> >> > Lars-Fredrik
> >> >> >
> >> >> >
> >> >> >
> >> >> >>
> >> >> >> > Regards
> >> >> >> > LF
> >> >> >> >
> >> >> >> > On Wed, Oct 15, 2014 at 8:48 AM, Romain Manni-Bucau <
> >> >> >> rmannibucau@gmail.com>
> >> >> >> > wrote:
> >> >> >> >
> >> >> >> >> Hi
> >> >> >> >>
> >> >> >> >> that's the global idea yes
> >> >> >> >>
> >> >> >> >> about transaction manager:
> >> >> >> >>
> >> >> >> >>
> >> >> >>
> >> >>
> >>
> http://grepcode.com/file/repo1.maven.org/maven2/org.apache.geronimo.components/geronimo-transaction/3.1.1/org/apache/geronimo/transaction/manager/TransactionManagerImpl.java#TransactionManagerImpl
> >> >> >> >> is th eone
> >> >> >> >>
> >> >> >> >> idea is simple: invoke registered hooks when needed
and just
> >> manage
> >> >> >> >> the lifecycle then EJB containers start/stop these
hooks
> calling
> >> the
> >> >> >> >> tx mgr (look SingletonContainer hierarchy in openejb-core
for
> >> >> >> >> instance)
> >> >> >> >>
> >> >> >> >>
> >> >> >> >> Romain Manni-Bucau
> >> >> >> >> @rmannibucau
> >> >> >> >> http://www.tomitribe.com
> >> >> >> >> http://rmannibucau.wordpress.com
> >> >> >> >> https://github.com/rmannibucau
> >> >> >> >>
> >> >> >> >>
> >> >> >> >> 2014-10-15 1:19 GMT+02:00 Lars-Fredrik Smedberg <
> >> itsmeden@gmail.com
> >> >> >:
> >> >> >> >> > Hi Romain
> >> >> >> >> >
> >> >> >> >> > Thanks for the link to the ManagedConnection,
some questions
> on
> >> it
> >> >> >> (so I
> >> >> >> >> > understand it correctly)
> >> >> >> >> >
> >> >> >> >> > - The connection is bound to the tx when the
first method is
> >> >> called on
> >> >> >> >> the
> >> >> >> >> > connect (except for toString, equals and hashcode),
and if
> its
> >> not
> >> >> >> >> already
> >> >> >> >> > bound, correct?
> >> >> >> >> > - Do I understand correctly that I will always
use (using
> >> >> delegation
> >> >> >> in
> >> >> >> >> the
> >> >> >> >> > proxy) the same connection (the one bound to
the tx)
> regardless
> >> of
> >> >> if
> >> >> >> I
> >> >> >> >> use
> >> >> >> >> > getConnection()/close() multiple times within
an EJB that
> >> handles
> >> >> tx?
> >> >> >> >> > - In the above case the connection I think I'm
using (when I
> >> simple
> >> >> >> look
> >> >> >> >> at
> >> >> >> >> > the code) is returned to the pool and the bound
one is used?
> >> >> >> >> >
> >> >> >> >> > Where can I get more understanding of the TransactionManager
> and
> >> >> its
> >> >> >> >> > relation to the transactions annotated in the
EJB?
> >> >> >> >> >
> >> >> >> >> > Very nice to take a look at the code, really
makes the
> >> >> understanding
> >> >> >> alot
> >> >> >> >> > easier, thanks!
> >> >> >> >> >
> >> >> >> >> > /LF
> >> >> >> >> >
> >> >> >> >> >
> >> >> >> >> > On Wed, Oct 15, 2014 at 1:03 AM, Romain Manni-Bucau
<
> >> >> >> >> rmannibucau@gmail.com>
> >> >> >> >> > wrote:
> >> >> >> >> >
> >> >> >> >> >> Hi
> >> >> >> >> >>
> >> >> >> >> >> this is actually more bound to JTA (you
can read
> >> >> >> >> >>
> >> >> >> >> >>
> >> >> >> >>
> >> >> >>
> >> >>
> >>
> http://svn.apache.org/repos/asf/tomee/tomee/trunk/container/openejb-core/src/main/java/org/apache/openejb/resource/jdbc/managed/local/ManagedConnection.java
> >> >> >> >> >> )
> >> >> >> >> >>
> >> >> >> >> >>
> >> >> >> >> >> 2014-10-15 0:14 GMT+02:00 Lars-Fredrik Smedberg
<
> >> >> itsmeden@gmail.com
> >> >> >> >:
> >> >> >> >> >> > Hi
> >> >> >> >> >> >
> >> >> >> >> >> > Using JDBC, creating a connection and
setting it to auto
> >> commit
> >> >> >> false
> >> >> >> >> to
> >> >> >> >> >> be
> >> >> >> >> >> > able to group more than one statement
into a single
> >> transaction
> >> >> I
> >> >> >> >> >> > understand.
> >> >> >> >> >> >
> >> >> >> >> >> > - Is it possible to compare that to
an EJB with a
> >> >> >> >> @TransactionAttribute
> >> >> >> >> >> > set? I would like to understand more
in detail how the EJB
> >> >> handles
> >> >> >> >> >> > transactions, datasources and connections
"under the
> hood".
> >> >> >> >> >> > (it all comes from a discussion with
a DBA)
> >> >> >> >> >> >
> >> >> >> >> >> > If I use @Resource to inject a Datasource
and use to
> >> retrieve a
> >> >> >> >> >> connection
> >> >> >> >> >> > I usually call close on it after I
done using it,
> >> >> >> >> >> >
> >> >> >> >> >> > - Will close on the connection actually
close it or is
> that
> >> >> done at
> >> >> >> >> the
> >> >> >> >> >> > commit of the transaction (assume we
are in an EJB with
> >> >> >> >> >> > @TransactionAttribute set)?
> >> >> >> >> >>
> >> >> >> >> >> will be done later to support rollback
> >> >> >> >> >>
> >> >> >> >> >> > - Will the container automatically
set auto commit false
> to
> >> any
> >> >> >> >> >> connections
> >> >> >> >> >> > retrieved within the transaction?
> >> >> >> >> >>
> >> >> >> >> >> all JTA ones, can also depend on JPA provider
config
> >> >> >> >> >>
> >> >> >> >> >> > - ... and many more questions :)
> >> >> >> >> >> >
> >> >> >> >> >> > Maybe I'm mixing the topics a bit but
it would be really
> >> great
> >> >> to
> >> >> >> >> >> > understand it all more in-depth to
be able to have the
> >> >> >> understanding
> >> >> >> >> and
> >> >> >> >> >> > discuss it with the DBA.
> >> >> >> >> >> >
> >> >> >> >> >> > Thanks
> >> >> >> >> >> >
> >> >> >> >> >> >
> >> >> >> >> >> >
> >> >> >> >> >> > --
> >> >> >> >> >> > Med vänlig hälsning / Best regards
> >> >> >> >> >> >
> >> >> >> >> >> > Lars-Fredrik Smedberg
> >> >> >> >> >> >
> >> >> >> >> >> > STATEMENT OF CONFIDENTIALITY:
> >> >> >> >> >> > The information contained in this electronic
message and
> any
> >> >> >> >> >> > attachments to this message are intended
for the exclusive
> >> use
> >> >> of
> >> >> >> the
> >> >> >> >> >> > address(es) and may contain confidential
or privileged
> >> >> >> information. If
> >> >> >> >> >> > you are not the intended recipient,
please notify
> >> Lars-Fredrik
> >> >> >> >> Smedberg
> >> >> >> >> >> > immediately at itsmeden@gmail.com,
and destroy all
> copies of
> >> >> this
> >> >> >> >> >> > message and any attachments.
> >> >> >> >> >>
> >> >> >> >> >
> >> >> >> >> >
> >> >> >> >> >
> >> >> >> >> > --
> >> >> >> >> > Med vänlig hälsning / Best regards
> >> >> >> >> >
> >> >> >> >> > Lars-Fredrik Smedberg
> >> >> >> >> >
> >> >> >> >> > STATEMENT OF CONFIDENTIALITY:
> >> >> >> >> > The information contained in this electronic
message and any
> >> >> >> >> > attachments to this message are intended for
the exclusive
> use
> >> of
> >> >> the
> >> >> >> >> > address(es) and may contain confidential or
privileged
> >> >> information. If
> >> >> >> >> > you are not the intended recipient, please notify
> Lars-Fredrik
> >> >> >> Smedberg
> >> >> >> >> > immediately at itsmeden@gmail.com, and destroy
all copies of
> >> this
> >> >> >> >> > message and any attachments.
> >> >> >> >>
> >> >> >> >
> >> >> >> >
> >> >> >> >
> >> >> >> > --
> >> >> >> > Med vänlig hälsning / Best regards
> >> >> >> >
> >> >> >> > Lars-Fredrik Smedberg
> >> >> >> >
> >> >> >> > STATEMENT OF CONFIDENTIALITY:
> >> >> >> > The information contained in this electronic message
and any
> >> >> >> > attachments to this message are intended for the exclusive
use
> of
> >> the
> >> >> >> > address(es) and may contain confidential or privileged
> >> information. If
> >> >> >> > you are not the intended recipient, please notify Lars-Fredrik
> >> >> Smedberg
> >> >> >> > immediately at itsmeden@gmail.com, and destroy all copies
of
> this
> >> >> >> > message and any attachments.
> >> >> >>
> >> >> >
> >> >> >
> >> >> >
> >> >> > --
> >> >> > Med vänlig hälsning / Best regards
> >> >> >
> >> >> > Lars-Fredrik Smedberg
> >> >> >
> >> >> > STATEMENT OF CONFIDENTIALITY:
> >> >> > The information contained in this electronic message and any
> >> >> > attachments to this message are intended for the exclusive use
of
> the
> >> >> > address(es) and may contain confidential or privileged
> information. If
> >> >> > you are not the intended recipient, please notify Lars-Fredrik
> >> Smedberg
> >> >> > immediately at itsmeden@gmail.com, and destroy all copies of this
> >> >> > message and any attachments.
> >> >>
> >> >
> >> >
> >> >
> >> > --
> >> > Med vänlig hälsning / Best regards
> >> >
> >> > Lars-Fredrik Smedberg
> >> >
> >> > STATEMENT OF CONFIDENTIALITY:
> >> > The information contained in this electronic message and any
> >> > attachments to this message are intended for the exclusive use of the
> >> > address(es) and may contain confidential or privileged information. If
> >> > you are not the intended recipient, please notify Lars-Fredrik
> Smedberg
> >> > immediately at itsmeden@gmail.com, and destroy all copies of this
> >> > message and any attachments.
> >>
> >
> >
> >
> > --
> > Med vänlig hälsning / Best regards
> >
> > Lars-Fredrik Smedberg
> >
> > STATEMENT OF CONFIDENTIALITY:
> > The information contained in this electronic message and any
> > attachments to this message are intended for the exclusive use of the
> > address(es) and may contain confidential or privileged information. If
> > you are not the intended recipient, please notify Lars-Fredrik Smedberg
> > immediately at itsmeden@gmail.com, and destroy all copies of this
> > message and any attachments.
>



-- 
Med vänlig hälsning / Best regards

Lars-Fredrik Smedberg

STATEMENT OF CONFIDENTIALITY:
The information contained in this electronic message and any
attachments to this message are intended for the exclusive use of the
address(es) and may contain confidential or privileged information. If
you are not the intended recipient, please notify Lars-Fredrik Smedberg
immediately at itsmeden@gmail.com, and destroy all copies of this
message and any attachments.

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message