[FLINK-3250] Savepoint coordinator requires too strict parallelism match

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 1.0.0
• Component/s: Runtime / Coordination
• Assignee: Ufuk Celebi
• Reporter: Ufuk Celebi
• Created: 18/Jan/16 12:55
• Updated: 18/Jan/16 16:50
• Resolved: 18/Jan/16 16:50

Description

The savepoint coordinator requires that the number of collected states of an operator and the parallelism of the new job (the one to be restored) match exactly. This is too strict.

For example a Kafka source with parallelism 2 and a single Kafka partition does not collect state for one of the two sources (hence it is not part of the savepoint state). Currently, restoring the same job with the same parallelism fails, which should not happen.