xiangfu0 opened a new issue, #20065: URL: https://github.com/apache/pinot/issues/20065
`ParquetAvroRecordReader.next()` wraps only `IOException` from `ParquetReader.read()`. In parquet-hadoop 1.18.1, unchecked `ParquetDecodingException` can escape; `SegmentMapper` classifies it as a parse error, bypassing `maxConsecutiveRecordFetchFailuresAllowed` when `continueOnError=true`. Reproduce by injecting a reader that returns the initial record and throws `ParquetDecodingException` on the next read, then calling `next()`. The fetch failure currently escapes without `RecordFetchException` classification. Wrap exceptions only around the prefetch read, matching the native reader, while preserving extraction failures as parse errors. Cover the retained cause, bounded termination at a configured fetch-failure threshold, and existing rejected-row/rewind behavior. This is a pre-existing follow-up requested in [the review of #20060](https://github.com/apache/pinot/pull/20060#discussion_r4396162588), separate from that PR's INT96 and cursor-progress changes. The narrow catch also exists at base `b58776bca94f1686b501025cd092c3104cc2cf33`. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
