[SPARK-32057] SparkExecuteStatementOperation does not set CANCELED state correctly

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0
• Fix Version/s: 3.0.1, 3.1.0
• Component/s: SQL
• Assignee: Ali Smesseim
• Reporter: Ali Smesseim
• Created: 22/Jun/20 21:00
• Updated: 12/Dec/22 18:11
• Resolved: 08/Jul/20 00:29

Description

https://github.com/apache/spark/pull/28671 introduced changes that changed the way cleanup is done in SparkExecuteStatementOperation. In cancel(), cleanup (killing jobs) used to be done after setting state to CANCELED. Now, the order is reversed. Jobs are killed first, causing exception to be thrown inside execute(), so the status of the operation becomes ERROR before being set to CANCELED.

cc Juliusz Sompolski

Issue Links

• is caused by: SPARK-31859 Thriftserver with spark.sql.datetime.java8API.enabled=true
• links to: [Github] Pull Request #28912 (alismess-db)
• links to: [Github] Pull Request #28912 (alismess-db)