slim bouguerra created HIVE-20875: ------------------------------------- Summary: Druid storage handler Kafka ingestion timestamp column name Key: HIVE-20875 URL: https://issues.apache.org/jira/browse/HIVE-20875 Project: Hive Issue Type: Task Reporter: slim bouguerra Assignee: Nishant Bangarwa
This question brought to my attention that currently the Druid-Hive-Kafka ingestion is assuming that the Kafka Stream has to include a column called __time as timestamp column. https://community.hortonworks.com/questions/226191/druid-kafka-ingestion-from-hive-hdp-30.html?childToView=227242#answer-227242 Looking at the code here seems confirming that https://github.com/apache/hive/blob/a51e6aeaf816bdeea5e91ba3a0fab8a31b3a496d/druid-handler/src/java/org/apache/hadoop/hive/druid/DruidStorageHandler.java#L301. IMO this is a serious limitation, because user can not always guarantee that the kafka record always will contain a column called `__time`, we need to introduce a have a way to configure this, maybe via table property. -- This message was sent by Atlassian JIRA (v7.6.3#76005)