[FLINK-8977] End-to-end test: Manually resume job after terminal failure

• Type: Sub-task
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Affects Version/s: 1.5.0
• Fix Version/s: 1.5.1, 1.6.0
• Component/s: Tests
• Assignee: Tzu-Li (Gordon) Tai
• Reporter: Till Rohrmann
• Created: 16/Mar/18 10:59
• Updated: 22/May/18 09:00
• Resolved: 22/May/18 09:00

Description

We should add an end-to-end test which verifies that a job can be resumed manually after a terminal job failure if there is a checkpoint. In order to do that we should

1. run the general purpose testing job FLINK-8971
2. wait for the completion of a checkpoint
3. Trigger a failure which leads to a terminal failure
4. Resume the job from the retained checkpoint

This end-to-end test should run with all state backend combinations: RocksDB (incremental/full, async/sync), FsStateBackend (async/sync).

Issue Links

• links to: GitHub Pull Request #6004