It sounds to me like you're not alway returning the
connection to the pool or not closing the connection.
But this is hard to say without the source of your
java class.
If you use a connection pool where you have
access to the source code, look if the get and release
methods can be intrumented in a way to log each
get and release with the hashcode of the connection.
Otherwise log each call to get/release in your own code
or build a wrapper object around the pool that log
this information before calling the pool method.
> -----Ursprüngliche Nachricht-----
> Von: Lukas Sägesser [mailto:[EMAIL PROTECTED]]
> Gesendet: Donnerstag, 21. Juni 2001 16:38
> An: [EMAIL PROTECTED]
> Betreff: mysteriously dying connections (Oracle <-> tomcat)
>
>
> hi all!
>
> problem:
> --------
>
> database access is managed via a java class that is
> instantiated and loaded into each clients session. every
> PL/SQL function and/or SQL statement a client needs is
> called in a method of this class.
> first, a connection is opened, the statement is called,
> all resultsets and statements are closed and finally the
> connection is closed. (or returned back to the
> connection pool if one is used)
>
> the driver we use is the Oracle jdbc ThinDriver (jdbc
> driver type 4). after running the system for several
> days, dead connections (and as we just discovered, tons
> of open cursors) pile up.
> eventually the webserver(tomcat) will crash or just
> hang.
> in the process, access gets slower and slower, there may
> be delays of 3 minutes until the webserver (or, perhaps
> the database, in which case the webserver would be just
> waiting for the db..) gives a response.
> after a restart of the webserver or the database or
> both, the system runns very well again.
>
> we tried several connection pools. with these, there
> were no delays for getting a connection, but the pool
> threw timeoutExceptions (i.e. made it visible that there
> are hanging connections) and after a few hundred
> timeouts the VM crashed with a stack overflow.
>
> the connections seem to get killed/kill themselfes
> randomly no matter which function was called. we log the
> methods and statements that were called for the
> connections that die, but there is just no pattern in
> it. thats why we dont find the root of the problem.. is
> it the driver ? the database ? the tomcat ?
>
>
>
> environment:
> ----------------
>
> Web Server:
> Suse Linux 7.2
> Apache Webserver /w Tomcat JSP Engine 3.2 /w SSL
>
> DB Server:
> Suse Linux 7.2
> Oracle 8i 8.1.7 /w multi-threaded server (MTS)
>
> Backup Server:
> Oracle Shadow server
>
>
>
> things done:
> ------------
>
> - linux update
> - apache update
> - tomcat update
> - jdbc thin driver update
> (everywhere the newest version)
> - tried several connection pools
>
>
>
> possible changes:
> -----------------
>
> - use OCI drivers instead of thin driver
> - use J-serv with GNU_jsp instead of tomcat
>
>
>
>
> every help is welcome, we're frustrated to no ends.
> thanks very much in advance!!
>
> l.sägesser
>
>
