[SPARK-49688] Data races detected by SparkConnectServiceSuite over ExecuteHolder

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 4.0.0
• Fix Version/s: 4.0.0
• Component/s: Connect
• Labels: pull-request-available
• Assignee: Changgyoo Park
• Reporter: Changgyoo Park
• Created: 17/Sep/24 15:47
• Updated: 24/Sep/24 23:58
• Resolved: 24/Sep/24 23:58

Description

1. Test issue.

https://github.com/longvu-db/spark/actions/runs/10903815379/job/30258793280

2. Real issue.

org.scalatest.exceptions.TestFailedDueToTimeoutException: The code passed to eventually never returned normally. Attempted 654 times over 10.001713506999998 seconds. Last failure message: Canceled did not equal Closed Execution has not been completed in Timeout(Span(10, Seconds)).

Data race among ExecuteThreadRunner, and interrupt.
1. Thread 1 starts an ExecuteHolder.
2. Thread 2 interrupts the ExecuteHolder, before 1 is completed.
3. The ExecuteHolder never reaches 'closed'.
=> Confirmed, reproduced.

Issue Links

• links to: GitHub Pull Request #48142
• links to: GitHub Pull Request #48208