[SPARK-20640] Make rpc timeout and retry for shuffle registration configurable

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.0.2
• Fix Version/s: 2.3.0
• Component/s: Shuffle, Spark Core
• Assignee: Li Yichao
• Reporter: Sital Kedia
• Created: 08/May/17 17:39
• Updated: 17/May/20 18:30
• Resolved: 21/Jun/17 13:55

Description

Currently the shuffle service registration timeout and retry has been hardcoded (see https://github.com/sitalkedia/spark/blob/master/network/shuffle/src/main/java/org/apache/spark/network/shuffle/ExternalShuffleClient.java#L144 and https://github.com/sitalkedia/spark/blob/master/core/src/main/scala/org/apache/spark/storage/BlockManager.scala#L197). This works well for small workloads but under heavy workload when the shuffle service is busy transferring large amount of data we see significant delay in responding to the registration request, as a result we often see the executors fail to register with the shuffle service, eventually failing the job. We need to make these two parameters configurable.

Issue Links

• is duplicated by: SPARK-20956 External shuffle server timeout
• links to: [Github] Pull Request #18092 (liyichao)