[ https://issues.apache.org/jira/browse/KYLIN-3393?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16498889#comment-16498889 ]
Billy Liu commented on KYLIN-3393: ---------------------------------- [~wuyingjun], Kylin does not depend on the Kafka Server, but the Kafka Client. So you do not need to install Kafka into kylin node, but copy the necessary client library. That's the same thing we are doing with Hadoop/Hive/HBase as Shaofeng said. Kylin keeps the dependency libraries out of the package to make it more compatible with the environment. If Kylin ships the fixed-version client libraries in binary, you have to recompile everything if the client library has conflicts with the server side. > Building streaming cube do not need to set “KAFKA_HOME” environment variable > ----------------------------------------------------------------------------- > > Key: KYLIN-3393 > URL: https://issues.apache.org/jira/browse/KYLIN-3393 > Project: Kylin > Issue Type: Improvement > Reporter: wuyingjun > Assignee: wuyingjun > Priority: Major > Attachments: 1.png, 2.png, KYLIN-3393.patch > > > You need hava KAFKA environment and set “KAFKA_HOME” environment variable > when building streaming cube, or else you can not submit a streaming cube > building job, like pic 1; > After we change the kafka dependency scope mode described in the > KYLIN-3393.patch , we can build streaming cube without a kafka environment in > the kylin service node successfully, like pic 2; -- This message was sent by Atlassian JIRA (v7.6.3#76005)