Mailing-List: contact commons-user-help@jakarta.apache.org; run by ezmlm
Precedence: bulk
Reply-To: "Jakarta Commons Users List" <commons-user@jakarta.apache.org>
Received-SPF: pass (asf.osuosl.org: domain of rjadrian@nc.rr.com designates
 24.25.9.103 as permitted sender)
Message-ID: <001901c5cb95$c59ec710$030aa8c0@teamreblaptop>
Reply-To: "Richard Road Runner" <rjadrian@nc.rr.com>
From: "Richard Road Runner" <rjadrian@nc.rr.com>
To: <commons-user@jakarta.apache.org>
Subject: DBCP connection pooling issue?
Date: Fri, 7 Oct 2005 19:20:56 -0400
MIME-Version: 1.0
Content-Type: multipart/alternative;
	boundary="----=_NextPart_000_0016_01C5CB74.3DF13530"

------=_NextPart_000_0016_01C5CB74.3DF13530
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

I am posting this issue on Sybase, commons and Tomcat newsgroups.  We =
are not sure right now which is responsible for our issue.

For a number of years now, we have been running into problems with the =
number of database connections growing unexpectedly.

We are running a 150 GB Sybase ASA 7 database..

We have a Java application running on Tomcat that uses commons DBCP.

Over time, the number of database connections increases and the database =
server eventually starts refusing connections.  We then need to restart =
the database server.

We have already explored the possiblity that we weren't closing =
connections properrly, however this does not appear to be the issue. =20

Another person blamed it on DBCP.  Below are two opinions.  One from =
this individual and the second from my supervisor.

"I have an issue with 5.5.9 and DBCP shipped with it I found 2 weeks=20
ago.  I would describe my problem as DBCP is not "reusing database=20
connections when it should" but it instead creates a new one, until the=20
SQL server has too many connections (or your DBCP hits upper parameter=20
limits).

With TC and MySQL I can confirm this by logging my database handle close =

and watching what happens with "SHOW PROCESSLIST" sql command on the=20
server.   Each request needs 1 database connection and the logging=20
should confirm it returns (releases) the connection back to DBCP pool. =20
I have followed this path with the debugger and got inside DBCP to prove =

the release takes place.

I had put down the problem and just configured up my DBCP parameters to=20
expire old connections as fast as possible and up the maximum limits. =20
This wont work for me in production, maybe I switch to C3P0 connection=20
pooling at that time.

Its just so difficult to help out and nail the problem with TCs=20
refactored and repackaged DBCP, I would very much appreciate an SDK=20
version of the full TC package that includes all source in the JARs.  I=20
am doing a CVS build.xml now, but I think this will be for 5.5.12, I=20
just need to work out how to fix the version for a 5.5.9 release and=20
compare the two trees to maybe try and nail this one, or at least=20
confirm the blame here is with DBCP.
"

Below is my supervisor's opionion. =20

"
Having JDBC work with an ASA database means enabling "java" in the =
database, and the database installs an (old) JRE on the server and =
presumably enables Java extensions from the sybase engine. There are =
patches at Sybase that upgrade the Sybase engine, but I bet the Java =
extensions lag behind the database upgrades. In addition, when you =
upgrade the JConnect driver in ORview, as I did fromJconnect 4.5 -> 5 =
->5.5, the new functionality in JConnect needs associated new =
functionality in the Sybase java extensions on the server in order to =
work fully. There are patches for serverside JConnect that I coulddn't =
install. The lag behing the JConnect on the client side and the Java =
extensions of our ASA server is my bet on the reason for our =
accumulating connections."

Any feedback would be helpfull
------=_NextPart_000_0016_01C5CB74.3DF13530--