[BEAM-12685] Allow managed thread count in AvroIO

• Type: Improvement
• Status: Open
• Priority: P3
• Resolution: Unresolved
• Component/s: io-java-avro, io-java-files
• Assignee: Unassigned
• Reporter: Dylan Hercher
• Created: 29/Jul/21 19:51
• Updated: 04/Jun/22 20:56

Description

During execution, the `ReadAllViaFileBasedSource` runs ReShuffle and creates an un-grouped set of file range readers.  This can easily cause OOM issues when the number of groups changes as there is no limit to the number of concurrent file reads.

 

Using Reshuffle.viaRandomKeys.withNumBuckets instead will allow the same default behavior, but lets the user configure the number of readers as and when needed.

Issue Links

• links to: GitHub Pull Request #15246