[SPARK-38776] Flaky test: ALSSuite.'ALS validate input dataset'

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.3.0
• Fix Version/s: 3.3.0
• Component/s: MLlib
• Assignee: Dongjoon Hyun
• Reporter: Dongjoon Hyun
• Created: 03/Apr/22 06:56
• Updated: 03/Apr/22 20:09
• Resolved: 03/Apr/22 17:21

Description

• https://github.com/apache/spark/runs/5803714260?check_suite_focus=true

[info] ALSSuite:
...
[info] - ALS validate input dataset *** FAILED *** (2 seconds, 449 milliseconds)
[info]   Invalid Long: out of range "Job aborted due to stage failure: Task 0 in stage 100.0 failed 1 times, most recent failure: Lost task 0.0 in stage 100.0 (TID 348) (localhost executor driver): org.apache.spark.SparkArithmeticException: Casting 1231000000000 to int causes overflow. To return NULL instead, use 'try_cast'. If necessary set spark.sql.ansi.enabled to false to bypass this error.

Issue Links

• is caused by: SPARK-38490 Add Github action test job for ANSI SQL mode
• links to: [Github] Pull Request #36051 (dongjoon-hyun)
• links to: [Github] Pull Request #36051 (dongjoon-hyun)
• links to: [Github] Pull Request #36054 (dongjoon-hyun)
• links to: [Github] Pull Request #36054 (dongjoon-hyun)