Using Confluent’s JDBC Connector without installing the entire platform

I was interested in trying out Confluent’s JDBC connector without installing their entire platform (I’d like to stick to vanilla Kafka as much as possible).  Here are the steps I followed to get it working with SQL Server.

Download Kafka 0.9, untar the archive, and create a directory named connect_libs in the kafka root (kafka_2.10-0.9.0.1/connect_libs).

Download the Confluent platform and extract the following jars (you should also be able to pull these from Confluent’s Maven repo, though I was unsuccessful):

• common-config-2.0.1.jar
• common-metrics-2.0.1.jar
• common-utils-2.0.1.jar
• kafka-connect-jdbc-2.0.1.jar

*Place these jars along with the SQL Server driver in kafka_2.10-0.9.0.1/connect_libs. Update bootstrap.servers in kafka_2.10-0.9.0.1/config/connect-standalone.properties with the broker list and create kafka_2.10-0.9.0.1/config/connect-jdbc.properties with the settings to try out:

name=sqlserver-feed
connector.class=io.confluent.connect.jdbc.JdbcSourceConnector
tasks.max=1

connection.url=jdbc:sqlserver://xxx.xxx.xxx.xxx:1433;databaseName=FeedDB;user=user;password=password
table.whitelist=tblFeedIP,tblFeedURL

mode=timestamp+incrementing
timestamp.column.name=localLastUpdated
incrementing.column.name=id

topic.prefix=stg-

Create the topics stg-tblFeedIP and stg-tblFeedURL on the cluster.

Add the connect_libs directory to the classpath:
export CLASSPATH="connect_libs/*"

And finally, run the connector in standalone mode with (make sure you are in the root kafka directory): bin/connect-standalone.sh config/connect-standalone.properties config/connect-jdbc.properties

Then, tail your topics to verify that messages are being produced by the connector.

* If you don’t care about cluttering up the default libs directory (kafka_2.10-0.9.0.1/libs), you can also just dump the jars there and not have to worry about setting the classpath.

Using Confluent’s JDBC Connector without installing the entire platform

标签：