Hi,
>>>The reason why this did not so much happen when the pooling
>>>is off may simply be that a connection is the closed, and thus
>>>the stuff is rolled back and released 'automatically'. You might
>>>have lost some updates in this case, however, but no lock
>>>persisting.
I don't switch the pooling of. I use a pooling class written by me.
My pool class maintains 45-60 connection and never closes a connection. It only
removes connections that are closed by the db server due to session timeouts and gets
a new connection for those. The dbtrace when using my connection pool does not show
any calls to get/setautocommit().
My app runs perfectly with this class.no updates are lost because all queries
are executed with autocommit set on by default. (I have checked this--> all
updates,deletes,inserts are immediately reflected in sqlstudio when using my
connection pool and by the way I never see a lock in the domain.locks either(assume
its due to autocommit)).The same application/same code works with both
oracle/postgreSQL JBOSS3.0 connection pooling so I'm sure something is broke in
SAPDB+JBOSS3.0.
I haven't been able to locate anyone else
using JBOSS3.0+SAPDB+connectionpooling.Also I noticed that Jboss.org does not have a
sapdb connection pool sample xml for JBOSS 3.0+.They only have a SAP DB connection
source.
What I don't understand is why the jboss transaction manager is setting
autocommit to false when executing some of my own queries and why it does not do this
consistently? (I make no request for beginning/ending a transaction.)
I would like to use jboss connection pool but will continue to use my own
connection pooling until someone finds a fix.
-Ajit
-----Original Message-----
From: Schroeder, Alexander [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, October 09, 2002 4:54 PM
To: ajit_cus; Sapdb. General@Listserv. Sap. Com
Subject: Locked Tables (was Bug in sap JDBC drivers and Jboss+SAP?)
Hi Ajis,
after looking at the trace I cannot see something weird, and/or unusual,
aside that there is never a 'commit' call in your servlet/bean which
will result in updates being kind of 'pending'.
So I assume there is a state originated in your application, which
results in some deadlock or similar phenomen.
I can only recommend here to make a 'SELECT * FROM LOCKSTATISTICS'
in case when this happens, and carefully look on the output
of this.
I think that the insert/update on your 'services' table will not be
committed, and thus the next read on the same row will see the
pending update and will not be able to establish the read lock
on that row.
The reason why this did not so much happen when the pooling
is off may simply be that a connection is the closed, and thus
the stuff is rolled back and released 'automatically'. You might
have lost some updates in this case, however, but no lock
persisting.
Regards
Alexander Schr�der
> -----Original Message-----
> From: ajit_cus [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, October 09, 2002 12:08 PM
> To: Schroeder, Alexander
> Subject: RE: Bug in sap JDBC drivers and Jboss+SAP
>
>
> Hi,
> I attach a sample database trace. Note that my application
> stalls and I can see row_exclusive locks in domain.locks
> using sqlstudio. The locks persist for a long time. Went
> through the trace and the autocommit setting is not always
> true when executing queries.
> Any assistance will be much appreciated.
> Thanks,
> Ajit
>
_______________________________________________
sapdb.general mailing list
[EMAIL PROTECTED]
http://listserv.sap.com/mailman/listinfo/sapdb.general
