danny0405 commented on issue #18844: URL: https://github.com/apache/hudi/issues/18844#issuecomment-4541910371
> prepareRecord clone produced by bufferInsertAndUpdate — i.e. populatedRecord.copy() after prependMetaFields. On the Spark engine these two objects are very different in heap size: > Incoming: HoodieSparkRecord wrapping a compact UnsafeRow (hundreds of bytes). Buffered (heap-retained): HoodieAvroIndexedRecord wrapping a fully-deserialized Avro IndexedRecord (many KB per record). The `HoodieSparkRecord.prependMetaFields` still returns a `HoodieSparkRecord` that holds an `UnsafeRow` in there, and the `HoodieSparkRecord.copy` still just clone the underneath `UnsafeRow`. Guess the major gap is the metadata fields there, and we can not ingore the delete records buffering size, the records are still buffered as `DeleteRecord`. -- 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]
