Definition

JDBC_PING uses a database to discover initial members. It was initially designed as an alternative to S3_PING, as some cloud deployed applications need access to a shared database anyway, so they could reuse credentials to access the database rather than deal with an additional service and relative authentication tokens. It uses a single table, so if that's not annoying you could share an existing Datasource with your application.

Connection configuration

Either provide all needed connection properties, or specify a JNDI name to lookup a Datasource in the environment.

Via connection properties:

Via JNDI lookup:

It's considered an illegal configuration to configure both JNDI related properties and plain connection related properties. It will attempt to create the needed table automatically.

Customize table definition and used SQL statements

The following properties permit you to customize all SQL related aspects, from changing the table create statement being used, disable automatic table creation, and the CRUD operations.

Update statements don't need to be atomic, and so they use a delete followed by an insert.

The default statements are:

CREATE TABLE JGROUPSPING (
   own_addr varchar(200) NOT NULL,
   cluster_name varchar(200) NOT NULL,
   ping_data varbinary(5000) DEFAULT NULL,
   PRIMARY KEY (own_addr, cluster_name) )

INSERT INTO JGROUPSPING (own_addr, cluster_name, ping_data) values (?, ?, ?)

DELETE FROM JGROUPSPING WHERE own_addr=? AND cluster_name=?

SELECT ping_data FROM JGROUPSPING WHERE cluster_name=?

Where own_addr and cluster_name are treated as String, while ping_data is a byte array. Keep consistency in the parameter order and choose compatible types appropriate to your database of choice.

Other configuration properties

The own Address is initially written at startup and deleted at shutdown, but it might be useful to have it rewrite it periodically by using the interval property.

Tested configurations

While the default SQL fragments use standard SQL in hope that they will be fine for you without needing reconfiguration, often specific databases require, or perform better, using different statements to take advantage of proprietary dialects.

MySQL recommended options

Oracle tested options

Configuration xml of JDBC_PING

<JDBC_PING datasource_jndi_name="java:/comp/env/sqlPool"

       initialize_sql="CREATE TABLE JGROUPSPING (

                       own_addr varchar(200) NOT NULL,

                       bind_addr varchar(200) NOT NULL,

                       created timestamp NOT NULL,

                       cluster_name varchar(200) NOT NULL,

                       ping_data blob,

                       constraint PK_JGROUPSPING PRIMARY KEY (own_addr, cluster_name)

                       )"

       insert_single_sql="INSERT INTO JGROUPSPING (own_addr, bind_addr, created, cluster_name, ping_data) values (?,

       '${jgroups.tcp.address:127.0.0.1}',sysdate, ?, ?)"

       delete_single_sql="DELETE FROM JGROUPSPING WHERE own_addr=? AND cluster_name=?"

       select_all_pingdata_sql="SELECT ping_data FROM JGROUPSPING WHERE cluster_name=?"

   />

  Additional attributes:

• BIND_ADDR is ip address to easily recognize which machine runs the node, jgroups.tcp.address  must be defined java property
• CREATED helps find which node joined later - detect possible partitioning in monitoring tools outside of running cluster ...