So,
the caching of the PreparedStatement is stored in the database
connection context and is not shared between the database connection. I am
thinking a work around way and it may solve the caching issue. If we
have the "named connection" from the pool with the lifecyclye control to release
the PreparedStatement from the applicaiton which invoke the container specific
API, it probably can solve the problem.
- Mike
Jau
-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 22, 2001 1:25 PM
To: [EMAIL PROTECTED]
Subject: RE: [JBoss-user] A little BMP philosophy/understanding
In the original JBoss 2.0 version the PreparedStatement cache was not discarded after the connection was returned to the pool because more than likely you might want to issue that one of these PreparedStatements again. To make matters worse there wasn't an upper limit on the number of PreparedStatement objects in the cache so things would continue to grow as you prepared new SQL statements. If you happened to prepare the same exact SQL statement then you received the previously cached PreparedStatement object but otherwise you got a new PreparedStatement that was also added to the cache. This would continue until either a) the database complained or b) you ran out of memory which ever came first. On Oracle, for example, each PreparedStatement takes memory on the database and once you hit 100 or so the database throws an exception when you try to get another one.
I patched the code by releasing the PreparedStatement cache when the Connection was released and submitted that fix but I'm not sure it was accepted. What really needs to happen is that the PreparedStatement cache needs to be enhanced so that an upper bound can be established via a configuration variable so that after x PreparedStatements have been cached new PreparedStatements will push one of the old ones out of the cache.
- Jon Harvie
Mike Jau <[EMAIL PROTECTED]>
Sent by: [EMAIL PROTECTED]03/22/2001 12:42 PM
Please respond to jboss-user
To: "'[EMAIL PROTECTED]'" <[EMAIL PROTECTED]>
cc:
Subject: RE: [JBoss-user] A little BMP philosophy/understanding
Could you give me some background information about the Preparedstaement
caching on the EJB container side?
Since the connection get from pool need to return to pool once the
transaction done. I assumed that the resouce associate to this connection
should be released and the released resoure include the preparedstatement.
Later on, the create preparedstatement will be invoked again from different
connection. How the preparedstatement cached is my question?
- Mike
-----Original Message-----
From: Bill Burke [mailto:[EMAIL PROTECTED]]
Sent: Thursday, March 22, 2001 12:10 PM
To: [EMAIL PROTECTED]
Subject: Re: [JBoss-user] A little BMP philosophy/understanding
Dan Christopherson wrote:
> On Fri, 23 Mar 2001, Peter Routtier-Wone wrote:
>
>>> Someone from this discussion group indicate that container might cache
the
>>> PreparedStatement.
>>
>> I can't speak with authority on this, but that rings true. I'm guessing
that
>> interception doesn't happen for the setEntityContext() method and
therefore
>> you actually create a PreparedStatement rather than receiving one from
the
>> pool.
>>
>>> Just for kicks, I gave it a try but transactions weren't completed and
>>> they'd just hang out there forever, blocking every other persistence and
>>> finder method until they timed out.
>>
>> That would bollox lifecycle management, and the described behaviour
wouldn't
>> be at all surprising.
>
> This is also a common bean bug: 'close()' should be called on every
> resultset, statement, and connection in a finally clause so that you know
> it happens every time.
>
>> On the other hand, I'd have thought that PreparedStatements would be far
>> less costly to manufacture than Connections, and therefore not worth the
>> overhead of managing a pool. I think I'll poke my nose into the source
and
>> see what's there.
>
> There's often communication with the database to create the
> PreparedStatement. That way it can pre-compile a query plan. There is a
> prepared statement cache in JBoss: in JBoss 2.0, it caused problems with
> Oracle's cursor limit (fixed in 2.1).
I'm re-writing the minerva PreparedStatement caching so it handles
cursor limit better. I'll submit the code tomorrow after I test it.
Bill
_______________________________________________
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
