[SPARK-39225] Support spark.history.fs.update.batchSize

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.4.0
• Fix Version/s: 3.4.0
• Component/s: Spark Core
• Labels: pull-request-available
• Assignee: Hai Tao
• Reporter: Hai Tao
• Created: 18/May/22 16:35
• Updated: 19/May/22 20:10
• Resolved: 19/May/22 20:09

Description

Current Spark History Server suffers when there are a large number of eventlog files under eventLog.dir: when a SHS starts, the initial scan may take a long time, and new eventlog files would not be scanned/parsed until the initial scan completes.

For example, if the initial scan takes 1-2 days(this is not uncommon in large environments), the newly finished spark jobs would not show up in SHS since their eventlog files are not scanned/parsed until the initial scan process finishes. This would result in a 1-2 days SHS malfunctioning since the newly finished spark jobs are most likely to be queried by users.

Issue Links

• links to: [Github] Pull Request #36597 (hai-tao-1)