[SPARK-39083] Fix FsHistoryProvider race condition between update and clean app data

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.2.1
• Fix Version/s: 3.3.0, 3.2.2
• Component/s: Web UI

Environment

CentOS Linux release 7.7.1908

openjdk version "1.8.0_262"
OpenJDK Runtime Environment (build 1.8.0_262-b10)
OpenJDK 64-Bit Server VM (build 25.262-b10, mixed mode)

• Assignee: Vu Tan
• Reporter: Vu Tan
• Created: 02/May/22 05:22
• Updated: 08/May/22 13:10
• Resolved: 08/May/22 13:10

Description

After SPARK-29043, FsHistoryProvider will list the log info without waiting all `mergeApplicationListing` task finished.

However, the method `cleanAppData` for stale data is not thread safe and it could lead to a scenario when `cleanAppData` delete the entry of 

ApplicationInfoWrapper for an application right after it has been updated by `mergeApplicationListing`.

So there will be cases when the HS Web UI displays `Application not found` for applications whose logs does exist.

Issue Links

• is duplicated by: SPARK-39082 Fix FsHistoryProvider race condition between update and clean app data
• relates to: SPARK-29043 [History Server]Only one replay thread of FsHistoryProvider work because of straggler
• relates to: SPARK-37659 Fix FsHistoryProvider race condition between list and delet log info
• links to: [Github] Pull Request #36424 (tanvn)