[FLINK-22255] AdaptiveScheduler improvements/bugs

• Type: Improvement
• Status: Open
• Priority: Minor
• Resolution: Unresolved
• Affects Version/s: 1.13.0
• Component/s: Runtime / Coordination
• Labels: auto-deprioritized-major pull-request-available
• Assignee: Unassigned
• Reporter: Till Rohrmann
• Created: 13/Apr/21 09:10
• Updated: 15/Jul/22 10:39

Description

This ticket collects the improvements/bugs for the AdaptiveScheduler.

Issue Links

• is related to: FLINK-21075 FLIP-160: Adaptive scheduler
• links to: GitHub Pull Request #18948

Sub-Tasks

• Only trigger scale up if the resources have stabilized: Open
• Reactive Mode parallelism changes are not shown in the job graph visualization in the UI: Closed
• Adaptive Scheduler: Show history of rescales in the Web UI: Open
• Checkpoint history lost across parallelism changes: Closed
• Introduce explicit shutdown signalling between TaskManager and JobManager: Resolved
• Introduce shared interface for adaptive scheduler state transitions: Closed
• Adaptive job shows error in WebUI when not enough resource are available: Closed