yihua opened a new issue, #20075: URL: https://github.com/apache/hudi/issues/20075
`HoodieSparkParquetReader` (native parquet log files and parquet log blocks) and `HoodieAvroParquetReader` build their `ParquetReader` from a path, which creates a new Hadoop `Configuration` and loads and parses the default resources for every file. Each parquet base file read through the Spark file group reader also copies the full Hadoop configuration three times. On executors with a large configuration this costs milliseconds of CPU per log file. Proposal: build the readers from a `HadoopInputFile` over the existing storage configuration, and copy the configuration once per base file read. part of #20064 -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
