[FLINK-35414] Cancel jobs through rest api for last-state upgrades

• Type: Improvement
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: kubernetes-operator-1.10.0
• Component/s: Kubernetes Operator
• Labels: pull-request-available
• Assignee: Gyula Fora
• Reporter: Gyula Fora
• Created: 21/May/24 14:51
• Updated: 07/Nov/24 23:51
• Resolved: 04/Oct/24 09:06

Description

The kubernetes operator currently always deletes the JM deployment directly during last-state upgrades instead of attempting any type of graceful shutdown.

We could improve the last-state upgrade logic to cancel the job in cases where the JM is healthy and then simply extract the last checkpoint info through the rest api like we already do for terminal job states.

This would allow the last-state upgrade mode to work even for session jobs and this may even eliminate a few corner cases that can result from the current forceful upgrade mechanism. 

Issue Links

• links to: GitHub Pull Request #871