[FLINK-22597] JobMaster cannot be restarted

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 1.11.3, 1.12.3
• Fix Version/s: 1.11.4, 1.12.4
• Component/s: Runtime / Coordination
• Labels: pull-request-available
• Assignee: Till Rohrmann
• Reporter: Till Rohrmann
• Created: 07/May/21 15:07
• Updated: 10/May/21 10:19
• Resolved: 10/May/21 10:19

Description

Since we are reusing the DefaultLeaderRetrievalService for finding the resource manager in the JobMaster and since the leader retrieval service cannot be reused, it is not possible to restart the JobMaster. This causes Flink to fail in case that the JobMaster regains the leadership in Flink versions <= 1.12. The problem no longer occurs for newer versions because of FLINK-11719.

Issue Links

• links to: GitHub Pull Request #15854
• links to: GitHub Pull Request #15855