DB2, HADR and Java

Introduction
HADR (High Availability and Disaster Recovery) is working well with C/C++ client but what about Java client. No problem, DB2 is also supporting Java without any problem for both DataSource and DriverManager client. It is described with details here.
Example for DriverManager client
I created a simple console Java application which connects to DB2 HADR controlled database and executes several JDBC action. Full source code is available here. How to setup a simple HADR configuration is described here.
Sample application description
There is a Java interfaces encapsulating some JDBC action.

public interface IJDBCAction {

        /**
         * Connect to database
         */
        public void connect();

        /**
         * Disconnect
         */
        public void disconnect();

        /**
         * Add new person to PERSON table.
         * 
         * @param re
         *            Person data to add (only first and family name should be
         *            filled)
         */
        public void addPerson(PersonRecord re);

        /**
         * Return the list of the persons
         * 
         * @return List of persons.
         */
        List<PersonRecord> getList();

        /**
         * Create PERSON table
         */
        public void createTable();

        /**
         * Drop PERSON table
         */
        public void dropTable();

        /**
         * Get current connection status
         * 
         * @return String describing the current status
         */
        public String getStatus();

        /**
         * Get the status of the last executed statement
         * 
         * @return String describing the status of the last statement
         */
        public String getLastResult();

        /**
         * Set autocommit mode
         * @param on true: set autocommit on, off otherwise
         */
        public void setAutocommit(boolean on);
}

There is also a concrete implementation of this interface, factory providing this implementation and simple console application with UI for this JDBC actions. Pay attention that it is the responsibility of the   IJDBCAction user to provide valid Connection reference. This way IJDBCAction implementation is free of any configuration details and can be reused in different context.
The application itself:

Connection status: Not connected
1) Connect
2) Disconnect
3) Read list
4) Add person
5) Create PERSON table
6) Drop PERSON table
7) Set autocommit on
8) Set autocommit off
99) Exit

There is nothing special in it, everything is a typical JDBC application. Because Java client does not use CLP connection pools, to make it HADR enabled connection URL string should be extended. It is described here.
The code sample used in the application.

  private static class GetConn implements IGetConnection {

                private final String user = "db2had1";
                private final String password = "db2had1";
                private final String url = "jdbc:db2://think:50001/sample";

                /** Parameters for HADR stand by, server name and port. */
                private final String serverAlternate = "think";
                private final int portAlternate = 50002;

                @Override
                public Connection getCon() throws SQLException, ClassNotFoundException {
                        Class.forName("com.ibm.db2.jcc.DB2Driver");
                        // Construct URL referencing HADR environment
                        String u = url
                                        + ":"
                                        + com.ibm.db2.jcc.DB2BaseDataSource.propertyKey_clientRerouteAlternateServerName
                                        + "=" + serverAlternate + ";";
                        u += com.ibm.db2.jcc.DB2BaseDataSource.propertyKey_clientRerouteAlternatePortNumber
                                        + "=" + portAlternate + ";";
                        // Connect !
                        Connection con = DriverManager.getConnection(u, user, password);
                        return con;
                }
        }

To make connection HADR enabled one has to add two additional parameters with server and port name for HADR stand by server and that is final. Nothing more is necessary.
Making the application HADR failover transparent
Extending URL string with HADR related parameters causes only that after HADR failover application is automatically redirected to another server. But current transaction is rollbacked with SQL error message thrown:

[jcc][t4][2027][11212][4.7.85] Użycie połączenia nie powiodło się, lecz połączenie to zostało ustanowione ponownie. Nazwa hosta lub adres IP to "think", a nazwa usługi lub numer portu to 50 001.
Ponowna próba ustawienia rejestrów specjalnych może, lecz nie musi zostać podjęta (kod przyczyny = 1). ERRORCODE=-4498, SQLSTATE=08506

The solution is to replay the last transaction. But the question could be raised - how to accomplish this task automatically without any user or db admin intervention, just make this problem completely transparent. There could be a lot of different ideas - the idea implements in this sample application is to use command design pattern. To run any JDBC related task one has to extend an abstract class:

/**
 * Command design pattern, abstract command
 * 
 * @author sbartkowski
 * 
 */
abstract class SqlCommand {

        protected final SQLCommandRunTime context;

        /** Error code 'connection reestablished' */
        private final int RETRYCODE = -4498;

        /**
         * Constructor
         * 
         * @param context
         *            SQLCommandRunTime context
         */
        SqlCommand(SQLCommandRunTime context) {
                this.context = context;
        }

        /**
         * Transactional command to be executed.
         * 
         * @throws SQLException
         *             SQLException thrown
         */
        abstract protected void command() throws SQLException;

        /**
         * Executes command, encapsulation around lower level transactional code.
         * Implements logic related to HADR failover
         */
        void runCommand() {
                if (context.con == null) {
                        context.lastResult = "Cannot execute, not connected";
                        return;
                }
                try {
                        command();
                        context.lastResult = "Command executed successfully";
                } catch (SQLException e) {
                        int errCode = e.getErrorCode();
                        if (errCode == RETRYCODE) {
                                // run again command in case of reestablish error
                                try {
                                        command();
                                        context.lastResult = "Command executed successfully";
                                } catch (SQLException e1) {
                                        // do not try again, throw exception unconditionally
                                        context.lastResult = e.getMessage();
                                        e.printStackTrace();
                                }
                        } else {
                                // throws exception if error code is not reestablish error
                                context.lastResult = e.getMessage();
                                e.printStackTrace();
                        }

                }
        }
}

This way all logic related to 'connection reestablished' SQL error is implemented in one place. Pay attention that command is replayed only once, second execution of the transaction should be completed without any problem. If it does not happen then it is much better to raise an exception at once without any delay.
Conclusion

• It is quite simple to enable Java application for HADR, only simple URL string extension is required.
• This sample application keeps all connection details hardcoded in the source code. In the most business application database connection details are reconfigurable outside the source code but by all means no problem also to extend it to HADR related parameters.
• This example is related only to connection extracted from DriverManager. DataSource connection (typical for web application) should be enabled for HADR another way.
• Making application HADR failover transparent requires proper handling of "A connection failed but has been reestablished." error exception. It is quite simple if application is started from the very beginning (for example by using the solution described above) but could be complicated in case of legacy code.