JNDI Datasource HOW-TO

&project; Les Hughes David Haraburda Glenn Nielsen Yoav Shapira JNDI Datasource HOW-TO

JNDI Datasource configuration is covered extensively in the JNDI-Resources-HOWTO. However, feedback from tomcat-user has shown that specifics for individual configurations can be rather tricky.

Here then are some example configurations that have been posted to tomcat-user for popular databases and some general tips for db usage.

You should be aware that since these notes are derived from configuration and/or feedback posted to tomcat-user YMMV :-). Please let us know if you have any other tested configurations that you feel may be of use to the wider audience, or if you feel we can improve this section in anyway.

Please note that JNDI resource configuration changed somewhat between Tomcat 5.0.x and Tomcat 5.5.x. You will most likely need to modify older JNDI resource configurations to match the syntax in the example below in order to make them work in Tomcat 7.x.x.

Also, please note that JNDI DataSource configuration in general, and this tutorial in particular, assumes that you have read and understood the Context and Host configuration references, including the section about Automatic Application Deployment in the latter reference.

java.sql.DriverManager supports the service provider mechanism. This feature is that all the available JDBC drivers that announce themselves by providing a META-INF/services/java.sql.Driver file are automatically discovered, loaded and registered, relieving you from the need to load the database driver explicitly before you create a JDBC connection. However, the implementation is fundamentally broken in all Java versions for a servlet container environment. The problem is that java.sql.DriverManager will scan for the drivers only once.

The JRE Memory Leak Prevention Listener that is included with Apache Tomcat solves this by triggering the drivers scan during Tomcat startup. This is enabled by default. It means that only libraries visible to the listener such as the ones in $CATALINA_BASE/lib will be scanned for database drivers. If you are considering disabling this feature, note that the scan would be triggered by the first web application that is using JDBC, leading to failures when this web application is reloaded and for other web applications that rely on this feature.

Thus, the web applications that have database drivers in their WEB-INF/lib directory cannot rely on the service provider mechanism and should register the drivers explicitly.

The list of drivers in java.sql.DriverManager is also a known source of memory leaks. Any Drivers registered by a web application must be deregistered when the web application stops. Tomcat will attempt to automatically discover and deregister any JDBC drivers loaded by the web application class loader when the web application stops. However, it is expected that applications do this for themselves via a ServletContextListener.

The default database connection pool implementation in Apache Tomcat relies on the libraries from the Apache Commons project. The following libraries are used:

Commons DBCP
Commons Pool

These libraries are located in a single JAR at $CATALINA_HOME/lib/tomcat-dbcp.jar. However, only the classes needed for connection pooling have been included, and the packages have been renamed to avoid interfering with applications.

DBCP 1.4 provides support for JDBC 4.0.

See the DBCP documentation for a complete list of configuration parameters.

A database connection pool creates and manages a pool of connections to a database. Recycling and reusing already existing connections to a database is more efficient than opening a new connection.

There is one problem with connection pooling. A web application has to explicitly close ResultSet's, Statement's, and Connection's. Failure of a web application to close these resources can result in them never being available again for reuse, a database connection pool "leak". This can eventually result in your web application database connections failing if there are no more available connections.

There is a solution to this problem. The Apache Commons DBCP can be configured to track and recover these abandoned database connections. Not only can it recover them, but also generate a stack trace for the code which opened these resources and never closed them.

To configure a DBCP DataSource so that abandoned database connections are removed and recycled add the following attribute to the Resource configuration for your DBCP DataSource:

removeAbandoned="true"

When available database connections run low DBCP will recover and recycle any abandoned database connections it finds. The default is false.

Use the removeAbandonedTimeout attribute to set the number of seconds a database connection has been idle before it is considered abandoned.

removeAbandonedTimeout="60"

The default timeout for removing abandoned connections is 300 seconds.

The logAbandoned attribute can be set to true if you want DBCP to log a stack trace of the code which abandoned the database connection resources.

logAbandoned="true"

The default is false.

0. Introduction

Versions of MySQL and JDBC drivers that have been reported to work:

MySQL 3.23.47, MySQL 3.23.47 using InnoDB,, MySQL 3.23.58, MySQL 4.0.1alpha
Connector/J 3.0.11-stable (the official JDBC Driver)
mm.mysql 2.0.14 (an old 3rd party JDBC Driver)

Before you proceed, don't forget to copy the JDBC Driver's jar into $CATALINA_HOME/lib.

1. MySQL configuration

Ensure that you follow these instructions as variations can cause problems.

Create a new test user, a new database and a single test table. Your MySQL user must have a password assigned. The driver will fail if you try to connect with an empty password.

mysql> GRANT ALL PRIVILEGES ON *.* TO javauser@localhost -> IDENTIFIED BY 'javadude' WITH GRANT OPTION; mysql> create database javatest; mysql> use javatest; mysql> create table testdata ( -> id int not null auto_increment primary key, -> foo varchar(25), -> bar int);

Note: the above user should be removed once testing is complete!

Next insert some test data into the testdata table.

mysql> insert into testdata values(null, 'hello', 12345); Query OK, 1 row affected (0.00 sec) mysql> select * from testdata; +----+-------+-------+ | ID | FOO | BAR | +----+-------+-------+ | 1 | hello | 12345 | +----+-------+-------+ 1 row in set (0.00 sec) mysql>

2. Context configuration

Configure the JNDI DataSource in Tomcat by adding a declaration for your resource to your Context.

For example:

3. web.xml configuration

Now create a WEB-INF/web.xml for this test application.

<web-app xmlns="http://java.sun.com/xml/ns/j2ee" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" xsi:schemaLocation="http://java.sun.com/xml/ns/j2ee http://java.sun.com/xml/ns/j2ee/web-app_2_4.xsd" version="2.4"> <description>MySQL Test App</description> <resource-ref> <description>DB Connection</description> <res-ref-name>jdbc/TestDB</res-ref-name> <res-type>javax.sql.DataSource</res-type> <res-auth>Container</res-auth> </resource-ref> </web-app>

4. Test code

Now create a simple test.jsp page for use later. <%@ taglib uri="http://java.sun.com/jsp/jstl/sql" prefix="sql" %> <%@ taglib uri="http://java.sun.com/jsp/jstl/core" prefix="c" %> <sql:query var="rs" dataSource="jdbc/TestDB"> select id, foo, bar from testdata </sql:query> <html> <head> <title>DB Test</title> </head> <body> <h2>Results</h2> <c:forEach var="row" items="${rs.rows}"> Foo ${row.foo}<br/> Bar ${row.bar}<br/> </c:forEach> </body> </html>

That JSP page makes use of JSTL's SQL and Core taglibs. You can get it from Apache Tomcat Taglibs - Standard Tag Library project — just make sure you get a 1.1.x release. Once you have JSTL, copy jstl.jar and standard.jar to your web app's WEB-INF/lib directory.

Finally deploy your web app into $CATALINA_BASE/webapps either as a warfile called DBTest.war or into a sub-directory called DBTest

Once deployed, point a browser at http://localhost:8080/DBTest/test.jsp to view the fruits of your hard work.

0. Introduction

Oracle requires minimal changes from the MySQL configuration except for the usual gotchas :-)

Drivers for older Oracle versions may be distributed as *.zip files rather than *.jar files. Tomcat will only use *.jar files installed in $CATALINA_HOME/lib. Therefore classes111.zip or classes12.zip will need to be renamed with a .jar extension. Since jarfiles are zipfiles, there is no need to unzip and jar these files - a simple rename will suffice.

For Oracle 9i onwards you should use oracle.jdbc.OracleDriver rather than oracle.jdbc.driver.OracleDriver as Oracle have stated that oracle.jdbc.driver.OracleDriver is deprecated and support for this driver class will be discontinued in the next major release.

1. Context configuration

In a similar manner to the mysql config above, you will need to define your Datasource in your Context. Here we define a Datasource called myoracle using the thin driver to connect as user scott, password tiger to the sid called mysid. (Note: with the thin driver this sid is not the same as the tnsname). The schema used will be the default schema for the user scott.

Use of the OCI driver should simply involve a changing thin to oci in the URL string.

2. web.xml configuration

You should ensure that you respect the element ordering defined by the DTD when you create you applications web.xml file.

<resource-ref> <description>Oracle Datasource example</description> <res-ref-name>jdbc/myoracle</res-ref-name> <res-type>javax.sql.DataSource</res-type> <res-auth>Container</res-auth> </resource-ref>

3. Code example

You can use the same example application as above (asuming you create the required DB instance, tables etc.) replacing the Datasource code with something like

Context initContext = new InitialContext(); Context envContext = (Context)initContext.lookup("java:/comp/env"); DataSource ds = (DataSource)envContext.lookup("jdbc/myoracle"); Connection conn = ds.getConnection(); //etc.

0. Introduction

PostgreSQL is configured in a similar manner to Oracle.

1. Required files

Copy the Postgres JDBC jar to $CATALINA_HOME/lib. As with Oracle, the jars need to be in this directory in order for DBCP's Classloader to find them. This has to be done regardless of which configuration step you take next.

2. Resource configuration

You have two choices here: define a datasource that is shared across all Tomcat applications, or define a datasource specifically for one application.

2a. Shared resource configuration

Use this option if you wish to define a datasource that is shared across multiple Tomcat applications, or if you just prefer defining your datasource in this file.

This author has not had success here, although others have reported so. Clarification would be appreciated here.

2b. Application-specific resource configuration

Use this option if you wish to define a datasource specific to your application, not visible to other Tomcat applications. This method is less invasive to your Tomcat installation.

Create a resource definition for your Context. The Context element should look something like the following.

3. web.xml configuration

<resource-ref> <description>postgreSQL Datasource example</description> <res-ref-name>jdbc/postgres</res-ref-name> <res-type>javax.sql.DataSource</res-type> <res-auth>Container</res-auth> </resource-ref>

4. Accessing the datasource

When accessing the datasource programmatically, remember to prepend java:/comp/env to your JNDI lookup, as in the following snippet of code. Note also that "jdbc/postgres" can be replaced with any value you prefer, provided you change it in the above resource definition file as well.

InitialContext cxt = new InitialContext(); if ( cxt == null ) { throw new Exception("Uh oh -- no context!"); } DataSource ds = (DataSource) cxt.lookup( "java:/comp/env/jdbc/postgres" ); if ( ds == null ) { throw new Exception("Data source not found!"); }

These solutions either utilise a single connection to the database (not recommended for anything other than testing!) or some other pooling technology.

Whilst not strictly addressing the creation of a JNDI DataSource using the OCI client, these notes can be combined with the Oracle and DBCP solution above.

In order to use OCI driver, you should have an Oracle client installed. You should have installed Oracle8i(8.1.7) client from cd, and download the suitable JDBC/OCI driver(Oracle8i 8.1.7.1 JDBC/OCI Driver) from otn.oracle.com.

After renaming classes12.zip file to classes12.jar for Tomcat, copy it into $CATALINA_HOME/lib. You may also have to remove the javax.sql.* classes from this file depending upon the version of Tomcat and JDK you are using.

Ensure that you have the ocijdbc8.dll or .so in your $PATH or LD_LIBRARY_PATH (possibly in $ORAHOME\bin) and also confirm that the native library can be loaded by a simple test program using System.loadLibrary("ocijdbc8");

You should next create a simple test servlet or jsp that has these critical lines:

DriverManager.registerDriver(new oracle.jdbc.driver.OracleDriver()); conn = DriverManager.getConnection("jdbc:oracle:oci8:@database","username","password");

where database is of the form host:port:SID Now if you try to access the URL of your test servlet/jsp and what you get is a ServletException with a root cause of java.lang.UnsatisfiedLinkError:get_env_handle.

First, the UnsatisfiedLinkError indicates that you have

a mismatch between your JDBC classes file and your Oracle client version. The giveaway here is the message stating that a needed library file cannot be found. For example, you may be using a classes12.zip file from Oracle Version 8.1.6 with a Version 8.1.5 Oracle client. The classeXXXs.zip file and Oracle client software versions must match.
A $PATH, LD_LIBRARY_PATH problem.
It has been reported that ignoring the driver you have downloded from otn and using the classes12.zip file from the directory $ORAHOME\jdbc\lib will also work.

Next you may experience the error ORA-06401 NETCMN: invalid driver designator

The Oracle documentation says : "Cause: The login (connect) string contains an invalid driver designator. Action: Correct the string and re-submit." Change the database connect string (of the form host:port:SID) with this one: (description=(address=(host=myhost)(protocol=tcp)(port=1521))(connect_data=(sid=orcl)))

Ed. Hmm, I don't think this is really needed if you sort out your TNSNames - but I'm not an Oracle DBA :-)

Here are some common problems encountered with a web application which uses a database and tips for how to solve them.

Tomcat runs within a JVM. The JVM periodically performs garbage collection (GC) to remove java objects which are no longer being used. When the JVM performs GC execution of code within Tomcat freezes. If the maximum time configured for establishment of a database connection is less than the amount of time garbage collection took you can get a database connection failure.

To collect data on how long garbage collection is taking add the -verbose:gc argument to your CATALINA_OPTS environment variable when starting Tomcat. When verbose gc is enabled your $CATALINA_BASE/logs/catalina.out log file will include data for every garbage collection including how long it took.

When your JVM is tuned correctly 99% of the time a GC will take less than one second. The remainder will only take a few seconds. Rarely, if ever should a GC take more than 10 seconds.

Make sure that the db connection timeout is set to 10-15 seconds. For the DBCP you set this using the parameter maxWait.

These can occur when one request gets a db connection from the connection pool and closes it twice. When using a connection pool, closing the connection just returns it to the pool for reuse by another request, it doesn't close the connection. And Tomcat uses multiple threads to handle concurrent requests. Here is an example of the sequence of events which could cause this error in Tomcat:

  Request 1 running in Thread 1 gets a db connection.

  Request 1 closes the db connection.

  The JVM switches the running thread to Thread 2

  Request 2 running in Thread 2 gets a db connection
  (the same db connection just closed by Request 1).

  The JVM switches the running thread back to Thread 1

  Request 1 closes the db connection a second time in a finally block.

  The JVM switches the running thread back to Thread 2

  Request 2 Thread 2 tries to use the db connection but fails
  because Request 1 closed it.

Here is an example of properly written code to use a database connection obtained from a connection pool:

  Connection conn = null;
  Statement stmt = null;  // Or PreparedStatement if needed
  ResultSet rs = null;
  try {
    conn = ... get connection from connection pool ...
    stmt = conn.createStatement("select ...");
    rs = stmt.executeQuery();
    ... iterate through the result set ...
    rs.close();
    rs = null;
    stmt.close();
    stmt = null;
    conn.close(); // Return to connection pool
    conn = null;  // Make sure we don't close it twice
  } catch (SQLException e) {
    ... deal with errors ...
  } finally {
    // Always make sure result sets and statements are closed,
    // and the connection is returned to the pool
    if (rs != null) {
      try { rs.close(); } catch (SQLException e) { ; }
      rs = null;
    }
    if (stmt != null) {
      try { stmt.close(); } catch (SQLException e) { ; }
      stmt = null;
    }
    if (conn != null) {
      try { conn.close(); } catch (SQLException e) { ; }
      conn = null;
    }
  }

Please note that although the above instructions place the JNDI declarations in a Context element, it is possible and sometimes desirable to place these declarations in the GlobalNamingResources section of the server configuration file. A resource placed in the GlobalNamingResources section will be shared among the Contexts of the server.

In order to get Realms to work, the realm must refer to the datasource as defined in the <GlobalNamingResources> or <Context> section, not a datasource as renamed using <ResourceLink>.