Oracle's DataSources don't do pooling in the sense of maintaining pools
of physical connections (nneded for speed). Oracle calls minerva's type
of pooling "connection caching" in their documentation for their JDBC
2.0 drivers.
Oracle provides an "example" implementation of a cache called
OracleConnectionCacheImpl (implements DataSource). You instantiate the
cache with one of the pooling data sources (OracleXADataSource or
OracleConnectionPoolDataSource). The cache then does what minerva does
in terms of pooling -- but not as conveniently or portably.
I also found what appears to be a limitation in the
OracleConnectionCacheImpl. It did not appear to work correctly with
OracleXADataSources. The cache allocates from the XA data source you
give it by calling getPooledConnection() -- not getXAConnection(). I
was able to call commit() on one of the connections from the supposedly
XAConnection cache, and that's not good. There was no source code
nearby, so I gave up and just went with minerva.
Now if I could just get an example minerva's JDBCDataSourceLoader
working. The trouble is that there are places that are outside the
container where you need data sources. One example is, I think, in
using the database for authentication. You need access to a connection,
but you are outside a container (in the SecurityInterceptor?). Another
example is in custom MBean services where you might want to use the
database to store configuration information.
You can use physical connections from the minerva XA pool outside the
container and manually commit them if you want, but it seems wrong to do
that.
Tim
Toby Allsopp wrote:
> On Tue, Apr 10, 2001 at 05:55:27PM +0200, Shahar Solomianik wrote:
>
>> Toby, I think you are right.
>> I replaced the mbean configuration to use XADataSourceLoader and minerva (I
>> had to patch minerva to support scrollable result sets...), tested the
>> server under load and it was MUCH faster.
>> I didnt get into details and log/trace files, but the differences in speed
>> were visible to the human eye ....
>
>
> I'd be very interested to hear of any deeper investigation of this. It
> might be as simple as the Oracle DataSource implementation not doing
> pooling.
>
> Toby.
>
> _______________________________________________
> JBoss-user mailing list
> [EMAIL PROTECTED]
> http://lists.sourceforge.net/lists/listinfo/jboss-user
_______________________________________________
JBoss-user mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/jboss-user
