GitHub user cshuo added a comment to the discussion: Use ZSTD as the default 
compression codec for Parquet files

> Can we just control this cleanly based on parquet write version alone.

@vinothchandar The leak occurs on the reader side, and ZSTD files written by 
newer versions may still be read by affected older profiles. But if you want 
use this change as a catalyst for engine upgrades, I think we can first control 
the default codec based on the writer profile. If community users later have a 
real need to read ZSTD files with affected older engine versions, we can 
introduce a scoped workaround at that point. WDYT?

BTW, the minimum parquet-java version used by the Flink profiles is 1.13.1, 
which contains the upstream fix, so Flink is not affected by the memory leak 
issue.

GitHub link: 
https://github.com/apache/hudi/discussions/19615#discussioncomment-18045856

----
This is an automatically sent email for [email protected].
To unsubscribe, please send an email to: [email protected]

Reply via email to