Kafka Connect Oracle

kafka-connect-oracle is a Kafka source connector for capturing all row based DML changes from Oracle database and streaming these changes to Kafka.Change data capture logic is based on Oracle LogMiner solution.

Only committed changes are pulled from Oracle which are Insert,Update,Delete operations. All streamed messages have related full "sql_redo" statement and parsed fields with values of sql statements. Parsed fields and values are kept in proper field type in schemas.

Messages have old (before change) and new (after change) values of row fields for DML operations.Insert operation has only new values of row tagged as "data",update operation has new data tagged as "data" and also contains old values of row before change tagged as "before".Delete operation only contains old data tagged as "before".

Sample data

Insert :

{    
    "SCN": 768889966828,
    "SEG_OWNER": "TEST",
    "TABLE_NAME": "TEST4",
    "TIMESTAMP": 1537958606000,
    "SQL_REDO": "insert into \"TEST\".\"TEST4\"(\"ID\",\"NAME\",\"PROCESS_DATE\",\"CDC_TIMESTAMP\") values (78238,NULL,NULL,TIMESTAMP ' 2018-09-26 10:43:26.643')",
    "OPERATION": "INSERT",
    "data": {
        "ID": 78238.0,
        "NAME": null,
        "PROCESS_DATE": null,
        "CDC_TIMESTAMP": 1537947806643
    },
    "before": null
}

Update :

{
    "SCN": 768889969452,
    "SEG_OWNER": "TEST",
    "TABLE_NAME": "TEST4",
    "TIMESTAMP": 1537964106000,
    "SQL_REDO": "update \"TEST\".\"TEST4\" set \"NAME\" = 'XaQCZKDINhTQBMevBZGGDjfPAsGqTUlCTyLThpmZ' where \"ID\" = 78238 and \"NAME\" IS NULL and \"PROCESS_DATE\" IS NULL and \"CDC_TIMESTAMP\" = TIMESTAMP ' 2018-09-26 10:43:26.643'",
    "OPERATION": "UPDATE",
    "data": {
        "ID": 78238.0,
        "NAME": "XaQCZKDINhTQBMevBZGGDjfPAsGqTUlCTyLThpmZ",
        "PROCESS_DATE": null,
        "CDC_TIMESTAMP": 1537947806643
    },
    "before": {
        "ID": 78238.0,
        "NAME": null,
        "PROCESS_DATE": null,
        "CDC_TIMESTAMP": 1537947806643
    }
}

Delete :

{
    "SCN": 768889969632,
    "SEG_OWNER": "TEST",
    "TABLE_NAME": "TEST4",
    "TIMESTAMP": 1537964142000,
    "SQL_REDO": "delete from \"TEST\".\"TEST4\" where \"ID\" = 78238 and \"NAME\" = 'XaQCZKDINhTQBMevBZGGDjfPAsGqTUlCTyLThpmZ' and \"PROCESS_DATE\" IS NULL and \"CDC_TIMESTAMP\" = TIMESTAMP ' 2018-09-26 10:43:26.643'",
    "OPERATION": "DELETE",
    "data": null,
    "before": {
        "ID": 78238.0,
        "NAME": "XaQCZKDINhTQBMevBZGGDjfPAsGqTUlCTyLThpmZ",
        "PROCESS_DATE": null,
        "CDC_TIMESTAMP": 1537947806643
    }
}

Setting Up

The database must be in archivelog mode and supplemental logging must be enabled.

On database server

sqlplus / as sysdba    
SQL>shutdown immediate
SQL>startup mount
SQL>alter database archivelog;
SQL>alter database open;

Enable supplemental logging

sqlplus / as sysdba    
SQL>alter database add supplemental log data (all) columns;

In order to execute connector successfully, connector must be started with privileged Oracle user.If given user has DBA role , this step can be skipped else following scripts will be executed to create privileged user.

create role logmnr_role;
grant create session to logmnr_role;
grant  execute_catalog_role,select any transaction ,select any dictionary to logmnr_role;
create user kminer identified by kminerpass;
grant  logmnr_role to kminer;
alter user kminer quota unlimited on users;

In a multitenant configuration, the privileged Oracle user must be a "common user" and some multitenant-specicic grants need to be made:

create role c##logmnr_role;
grant create session to c##logmnr_role;
grant  execute_catalog_role,select any transaction ,select any dictionary,logmining to c##logmnr_role;
create user c##kminer identified by kminerpass;
grant  c##logmnr_role to c##kminer;
alter user c##kminer quota unlimited on users set container_data = all container = current;

Configuration

Configuration Properties

Name	Type	Description
name	String	Connector name
connector.class	String	The name of the java class for this connector.
db.name.alias	String	Identifier name for database like Test,Dev,Prod or specific name to identify the database.This name will be used as header of topics and schema names.
tasks.max	Integer	Maximum number of tasks to create.This connector uses a single task.
topic	String	Name of the topic that the messages will be written to.If it is set a value all messages will be written into this declared topic , if it is not set, for each database table a topic will be created dynamically.
db.name	String	Service name or sid of the database to connect.Mostly database service name is used.
db.hostname	String	Ip address or hostname of Oracle database server.
db.port	Integer	Port number of Oracle database server.
db.user	String	Name of database user which is used to connect to database to start and execute logminer. This user must provide necessary privileges mentioned above.
db.user.password	String	Password of database user.
db.fetch.size	Integer	This config property sets Oracle row fetch size value.
table.whitelist	String	A comma separated list of database schema or table names which will be captured. For all schema capture <SCHEMA_NAME>.* For table capture <SCHEMA_NAME>.<TABLE_NAME> must be specified.
parse.dml.data	Boolean	If it is true , captured sql DML statement will be parsed into fields and values.If it is false only sql DML statement is published.
reset.offset	Boolean	If it is true , offset value will be set to current SCN of database when connector started.If it is false connector will start from last offset value.
start.scn	Long	If it is set , offset value will be set this specified value and logminer will start at this SCN.If connector would like to be started from desired SCN , this property can be used.
multitenant	Boolean	If true, multitenant support is enabled. If false, single instance configuration will be used.

Example Config

name=oracle-logminer-connector
connector.class=com.ecer.kafka.connect.oracle.OracleSourceConnector
db.name.alias=test
tasks.max=1
topic=cdctest
db.name=testdb
db.hostname=10.1.X.X
db.port=1521
db.user=kminer
db.user.password=kminerpass
db.fetch.size=1
table.whitelist=TEST.*,TEST2.TABLE2
parse.dml.data=true
reset.offset=false
multitenant=false

Building and Running

mvn clean package

Copy kafka-connect-oracle-1.0.jar and lib/ojdbc7.jar to KAFKA_HOME/lib folder. If CONFLUENT platform is using for kafka cluster , copy ojdbc7.jar and  kafka-connect-oracle-1.0.jar to $CONFLUENT_HOME/share/java/kafka-connect-jdbc folder.

Copy config/OracleSourceConnector.properties file to $KAFKA_HOME/config . For CONFLUENT copy properties file to $CONFLUENT_HOME/etc/kafka-connect-jdbc

To start connector

cd $KAFKA_HOME
./bin/connect-standalone.sh ./config/connect-standalone.properties ./config/OracleSourceConnector.properties

In order to start connector with Avro serialization 

cd $CONFLUENT_HOME

./bin/connect-standalone ./etc/schema-registry/connect-avro-standalone.properties ./etc/kafka-connect-jdbc/OracleSourceConnector.properties 

Do not forget to populate connect-standalone.properties  or connect-avro-standalone.properties with the appropriate hostnames and ports.

Todo:

1.Implementation for DDL operations
2.Support for other Oracle specific data types
3.New implementations
4.Performance Tuning
5.Initial data load
6.Bug fix