[FLINK-19632] Introduce a New ResultPartitionType for Approximate Local Recovery

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Component/s: Runtime / Task
• Labels: pull-request-available
• Assignee: Yuan Mei
• Reporter: Yuan Mei
• Created: 14/Oct/20 03:58
• Updated: 22/Jun/21 13:55
• Resolved: 02/Nov/20 08:35

Description

1. On downstream node failure, the upstream node needs to release a sub-partition view while keeping the result partition
2. When the downstream node reconnects, the subpartition view needs to be recreated. If the previous subpartition view still exists, it needs to be released.

Clean up of the partial records should be handled in a separate Jira ticket FLINK-19547

 

Issue Links

• links to: GitHub Pull Request #13648