[SPARK-19753] Remove all shuffle files on a host in case of slave lost of fetch failure

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.0.1
• Fix Version/s: 2.3.0
• Component/s: Scheduler, Spark Core
• Assignee: Sital Kedia
• Reporter: Sital Kedia
• Created: 27/Feb/17 20:31
• Updated: 17/May/20 17:46
• Resolved: 14/Jun/17 03:34

Description

Currently, when we detect fetch failure, we only remove the shuffle files produced by the executor, while the host itself might be down and all the shuffle files are not accessible. In case we are running multiple executors on a host, any host going down currently results in multiple fetch failures and multiple retries of the stage, which is very inefficient. If we remove all the shuffle files on that host, on first fetch failure, we can rerun all the tasks on that host in a single stage retry.

Issue Links

• is related to: SPARK-20178 Improve Scheduler fetch failures
• links to: [Github] Pull Request #17088 (sitalkedia)
• links to: [Github] Pull Request #18150 (sitalkedia)