You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@commons.apache.org by "Barclay A. Dunn" <ba...@happypuppy.com> on 2003/08/08 23:57:56 UTC
DBCP resource leaks
hi,
i am using a try / catch / finally block to close EVERY connection in my
code. i mean 100%. however, i am still seeing connections not getting
returned to the pool.
someone on the tomcat-users list gave me code (thanks angus) to display the
connection pool state - active and idle links. i can also log in to terminal
services client and watch my sql server's connections go up (and sometimes
down, but more up than down). so what i see is, the total number that sql
server thinks are open is more than the number that the pool says are open
(active and idle). and the totals are gradually increasingly divergent.
ultimately, it crashes tomcat.
can anybody suggest how i might go about finding the holes and plugging
them? if they are not in my code, where are they?
thanks,
barclay
RE: DBCP resource leaks
Posted by Michael Holly <mh...@talisentech.com>.
I had much the same problem. Starting/Stopping webapps did not deal with
cleaning up the pool correctly.
Since I was trying to implement junit and other testing mechanisims I wanted
my build script to stop, build, and
start the the app, then test it. My problem was that the pools were being
treated as a quasi server/webapp resource.
For the current configurations I couldn't find one that would take ownership
of the pool. Then somebody told me
to create a Context listener and use it to clean up the pool when the webapp
shutdown occurs. After several
iterations I go this to work. I have checked it against my db through
several shutdown startup cycles and
it has yet to lose track of a connection.
Here is my Context Listener
/**
* The listener runs when the app is started and shutdown
*
* @author Michael Holly
* created Apr 18, 2003
*/
package net.talisen.tsr;
import javax.sql.DataSource;
import javax.naming.InitialContext;
import javax.naming.Context;
import javax.naming.NamingException;
import javax.naming.NamingEnumeration;
import org.apache.commons.dbcp.BasicDataSource;
import javax.servlet.*;
import org.apache.log4j.Logger;
import java.util.ResourceBundle;
import java.net.URL;
import java.net.MalformedURLException;
import java.io.*;
import java.util.*;
import org.apache.log4j.Logger;
import org.apache.log4j.Level;
import org.apache.log4j.PropertyConfigurator;
public final class ContextListener
implements ServletContextListener
{
//get a logger
Logger log = Logger.getLogger(ContextListener.class);
private InitialContext initialContext = null;
private Context namingContext = null;
private ServletContext context = null;
public void contextInitialized (ServletContextEvent servletContextEvent)
{
context = servletContextEvent.getServletContext ();
try
{
System.out.println(" \n\n Intializing Context");
log.info("Initializing logging");
// configure the Log4j system
String file = new String( "/WEB-INF/classes/log4j.properties" );
URL url = context.getResource(file);
PropertyConfigurator.configure( url );
System.out.println("Log4j Properties @ " + url.toString() );
log.info("Cataloging Context Resources");
initialContext = new InitialContext();
namingContext = (Context) initialContext.lookup("java:comp/env");
DataSource ds1 =
(DataSource)namingContext.lookup("jdbc/oracle_myapp");
DataSource ds2 =
(DataSource)namingContext.lookup("jdbc/oracle_company_db");
context.setAttribute("dataSource1", ds1);
context.setAttribute("dataSource2", ds2);
log.info("oracle_myapp connection pool cataloged");
log.info("oracle_company_db connection pool cataloged");
}
catch (NamingException ne)
{
log.error("Couldn't create context attribute: " + ne.getMessage
());
ne.printStackTrace();
}
catch (Exception e)
{
log.error("Couldn't create context attribute: " + e.getMessage ());
e.printStackTrace();
}
}
public void contextDestroyed (ServletContextEvent servletContextEvent)
{
DataSource ds1 = ((DataSource) context.getAttribute("dataSource1"));
DataSource ds2 = ((DataSource) context.getAttribute("dataSource2"));
try
{
log.info("Cleaning up Context Resources");
if (ds1 instanceof org.apache.commons.dbcp.BasicDataSource) {
log.info("Found oracle_tsr connection pool " + ds1.toString());
((org.apache.commons.dbcp.BasicDataSource) ds1).close();
ds1 = null;
}
log.info("Removed oracle_myapp connection ");
if (ds2 instanceof org.apache.commons.dbcp.BasicDataSource) {
log.info("Found oracle_talisen connection pool " +
ds2.toString() );
((org.apache.commons.dbcp.BasicDataSource) ds2).close();
ds2 = null;
}
log.info("Removed oracle_company_db connection");
context.removeAttribute ("dataSource1");
context.removeAttribute ("dataSource2");
}
catch (Exception e)
{
log.error("Error destroying Context: " + e.getMessage ());
e.printStackTrace();
}
finally
{
System.out.println("########################################################
###########################################");
System.out.println("########################################################
###########################################");
System.out.println("");
System.out.println("");
System.out.println("");
System.out.println("");
System.out.println("");
System.out.println("");
System.out.println("");
System.out.println("");
}
}
}
then I add the Contect Listener to the web.xml file
<!-- CONTEXT LISTENER -->
<listener>
<listener-class>net.talisen.tsr.ContextListener</listener-class>
</listener>
My server.xml does not contain any of the config info for the pools. This
is
contained in a context xml file. This file get stuffed into the META-INF
dir.
Then the whole thing gets jared up. I deploy using Tomcat Manager.
<!--Context path="/tsr" docBase="/tsr" debug="9" reloadable="true"
crossContext="true"-->
<!--Context path="/tsr" docBase="tsr.war" debug="9" reloadable="true"
crossContext="true"-->
<Context path="/tsr" debug="9" privileged="true" docBase="C:\Tomcat
4.1\webapps\tsr" useNaming="true">
<Loader checkInterval="6"/>
<Logger className="org.apache.catalina.logger.FileLogger"
prefix="localhost_tsrdb_log."
suffix=".txt"
timestamp="true"/>
<Resource name="jdbc/oracle_tsr" auth="Container" scope="Shareable"
type="javax.sql.DataSource"/>
<ResourceParams name="jdbc/oracle_myapp">
<parameter>
<name>factory</name>
<value>org.apache.commons.dbcp.BasicDataSourceFactory</value>
</parameter>
<parameter>
<name>driverClassName</name>
<!--value>oracle.jdbc.OracleDriver</value-->
<!--value>oracle.jdbc.pool.OracleConnectionPoolDataSource</value-->
<value>oracle.jdbc.pool.OracleDataSource</value>
</parameter>
<parameter>
<name>url</name>
<!--value>jdbc:oracle:thin:@myserver:1521:myco</value-->
</parameter>
<parameter>
<name>username</name>
<value>myapp</value>
</parameter>
<parameter>
<name>password</name>
<value>????????</value>
</parameter>
<parameter>
<name>maxActive</name>
<value>20</value>
</parameter>
<parameter>
<name>maxIdle</name>
<value>10</value>
</parameter>
<parameter>
<name>maxWait</name>
<value>-1</value>
</parameter>
<parameter>
<name>removeAbandoned</name>
<value>true</value>
</parameter>
<parameter>
<name>removeAbandonedTimeout</name>
<value>300</value>
</parameter>
<parameter>
<name>logAbandoned</name>
<value>true</value>
</parameter>
<parameter>
<name>validationQuery</name>
<value>select 'validationQuery' from dual</value>
</parameter>
<parameter>
<name>testOnBorrow</name>
<value>true</value>
</parameter>
</ResourceParams>
<Resource name="jdbc/oracle_talisen" auth="Container"
type="javax.sql.DataSource"/>
<ResourceParams name="jdbc/oracle_talisen">
<parameter>
<name>factory</name>
<value>org.apache.commons.dbcp.BasicDataSourceFactory</value>
</parameter>
<parameter>
<name>driverClassName</name>
<value>oracle.jdbc.OracleDriver</value>
<!--value>oracle.jdbc.pool.OracleConnectionPoolDataSource</value-->
<!--value>oracle.jdbc.pool.OracleDataSource</value-->
</parameter>
<parameter>
<name>url</name>
<!--value>jdbc:oracle:thin:@myserver:1521:myco</value-->
</parameter>
<parameter>
<name>username</name>
<value>company_db</value>
</parameter>
<parameter>
<name>password</name>
<value>?????????</value>
</parameter>
<parameter>
<name>maxActive</name>
<value>20</value>
</parameter>
<parameter>
<name>maxIdle</name>
<value>10</value>
</parameter>
<parameter>
<name>maxWait</name>
<value>-1</value>
</parameter>
<parameter>
<name>removeAbandoned</name>
<value>true</value>
</parameter>
<parameter>
<name>removeAbandonedTimeout</name>
<value>300</value>
</parameter>
<parameter>
<name>logAbandoned</name>
<value>true</value>
</parameter>
<parameter>
<name>validationQuery</name>
<value>select 'validationQuery' from dual</value>
</parameter>
<parameter>
<name>testOnBorrow</name>
<value>true</value>
</parameter>
</ResourceParams>
</Context>
Hope this helps.
Michael
-----Original Message-----
From: Charles So [mailto:charles_so@mac.com]
Sent: Sunday, August 10, 2003 10:00 AM
To: Jakarta Commons Users List
Subject: Re: DBCP resource leaks
Hi,
Are you developing on the machine that you experience the problem? I
found out that if I leave Tomcat running , as in production, DBCP works
fine. However, if I start/stop/restart any webapp in Tomcat (such as in
development) then the previous opened connections will not be closed -
until i shutdown Tomcat and restart it.
Someone posted code about killing those connections when webapp is
restarted. Maybe you can see if that helps.
Hope this help.
Charles
On Sunday, August 10, 2003, at 10:03 PM, Dirk Verbeeck wrote:
> Hi Barclay
>
> Maybe you should first try a different JDBC driver.
> NetDirect is popular in the slide user group.
> See:
> http://faq.globalvision.com.au/cgi-bin/fom?_recurse=1&file=45#file_94
>
> Hope it helps
> Dirk
>
> Barclay A. Dunn wrote:
>
>> hi,
>>
>> i am using a try / catch / finally block to close EVERY connection in
>> my
>> code. i mean 100%. however, i am still seeing connections not getting
>> returned to the pool.
>>
>> someone on the tomcat-users list gave me code (thanks angus) to
>> display the
>> connection pool state - active and idle links. i can also log in to
>> terminal
>> services client and watch my sql server's connections go up (and
>> sometimes
>> down, but more up than down). so what i see is, the total number that
>> sql
>> server thinks are open is more than the number that the pool says are
>> open
>> (active and idle). and the totals are gradually increasingly
>> divergent.
>> ultimately, it crashes tomcat.
>>
>> can anybody suggest how i might go about finding the holes and
>> plugging
>> them? if they are not in my code, where are they?
>>
>> thanks,
>> barclay
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
>> For additional commands, e-mail: commons-user-help@jakarta.apache.org
>>
>>
>>
>>
>>
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-user-help@jakarta.apache.org
>
---------------------------------------------------------------------
To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
For additional commands, e-mail: commons-user-help@jakarta.apache.org
Re: DBCP resource leaks
Posted by Charles So <ch...@mac.com>.
Hi,
Are you developing on the machine that you experience the problem? I
found out that if I leave Tomcat running , as in production, DBCP works
fine. However, if I start/stop/restart any webapp in Tomcat (such as in
development) then the previous opened connections will not be closed -
until i shutdown Tomcat and restart it.
Someone posted code about killing those connections when webapp is
restarted. Maybe you can see if that helps.
Hope this help.
Charles
On Sunday, August 10, 2003, at 10:03 PM, Dirk Verbeeck wrote:
> Hi Barclay
>
> Maybe you should first try a different JDBC driver.
> NetDirect is popular in the slide user group.
> See:
> http://faq.globalvision.com.au/cgi-bin/fom?_recurse=1&file=45#file_94
>
> Hope it helps
> Dirk
>
> Barclay A. Dunn wrote:
>
>> hi,
>>
>> i am using a try / catch / finally block to close EVERY connection in
>> my
>> code. i mean 100%. however, i am still seeing connections not getting
>> returned to the pool.
>>
>> someone on the tomcat-users list gave me code (thanks angus) to
>> display the
>> connection pool state - active and idle links. i can also log in to
>> terminal
>> services client and watch my sql server's connections go up (and
>> sometimes
>> down, but more up than down). so what i see is, the total number that
>> sql
>> server thinks are open is more than the number that the pool says are
>> open
>> (active and idle). and the totals are gradually increasingly
>> divergent.
>> ultimately, it crashes tomcat.
>>
>> can anybody suggest how i might go about finding the holes and
>> plugging
>> them? if they are not in my code, where are they?
>>
>> thanks,
>> barclay
>>
>>
>> ---------------------------------------------------------------------
>> To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
>> For additional commands, e-mail: commons-user-help@jakarta.apache.org
>>
>>
>>
>>
>>
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
> For additional commands, e-mail: commons-user-help@jakarta.apache.org
>
Re: DBCP resource leaks
Posted by Dirk Verbeeck <di...@pandora.be>.
Hi Barclay
Maybe you should first try a different JDBC driver.
NetDirect is popular in the slide user group.
See: http://faq.globalvision.com.au/cgi-bin/fom?_recurse=1&file=45#file_94
Hope it helps
Dirk
Barclay A. Dunn wrote:
>hi,
>
>i am using a try / catch / finally block to close EVERY connection in my
>code. i mean 100%. however, i am still seeing connections not getting
>returned to the pool.
>
>someone on the tomcat-users list gave me code (thanks angus) to display the
>connection pool state - active and idle links. i can also log in to terminal
>services client and watch my sql server's connections go up (and sometimes
>down, but more up than down). so what i see is, the total number that sql
>server thinks are open is more than the number that the pool says are open
>(active and idle). and the totals are gradually increasingly divergent.
>ultimately, it crashes tomcat.
>
>can anybody suggest how i might go about finding the holes and plugging
>them? if they are not in my code, where are they?
>
>thanks,
>barclay
>
>
>---------------------------------------------------------------------
>To unsubscribe, e-mail: commons-user-unsubscribe@jakarta.apache.org
>For additional commands, e-mail: commons-user-help@jakarta.apache.org
>
>
>
>
>
>