[TEZ-4508] Allow the FAIR_PARALLELISM mode to accept multiple source vertices

• Type: Sub-task
• Status: Patch Available
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 0.10.2
• Assignee: Shohei Okumiya
• Reporter: Shohei Okumiya
• Created: 07/Aug/23 08:14
• Updated: 28/May/24 13:25

Description

Currently, FairShuffleVertexManager with FAIR_PARALLELISM fails when two or more edges are configured with SCATTER_GATHER.
https://github.com/apache/tez/blob/rel/release-0.10.2/tez-runtime-library/src/main/java/org/apache/tez/dag/library/vertexmanager/FairShuffleVertexManager.java#L198-L204

Looking at TEZ-3500, we assume the situation with such multiple sources happens when the vertex performs JOIN. In that case, I agree we definitely need more than the current FairShuffleVertexManager.

However, the current fair routing with multiple sources sufficiently works when the sources are symmetric. One case I assume is UNION ALL + bucketed INSERT.

Issue Links

• links to: GitHub Pull Request #306