mukul-8 commented on PR #27806:
URL: https://github.com/apache/flink/pull/27806#issuecomment-4161866251

   One thing to consider with this upgrade — Avro 1.12.1 enables 
fastReaderEnabled=true by default in GenericDatumReader/
   SpecificDatumReader. This introduces an optimized execution plan for 
deserialization that gets rebuilt whenever
   setSchema() is called.
   
   In Flink, AvroDeserializationSchema.deserialize() calls 
datumReader.setSchema(readerSchema) on every record. With the
   fast reader enabled, this means the optimized execution plan is rebuilt per 
record, which could negate the performance
   benefit and potentially cause a regression compared to 1.11.x.
   
   This is discussed in detail in #27499 
([FLINK-39005](https://issues.apache.org/jira/browse/FLINK-39005)). It would be 
worth studying the impact before merging — either benchmarking the current 
setSchema-per-record path with fastReaderEnabled=true, or explicitly setting 
fastReaderEnabled=false until the redundant setSchema() call is addressed.


-- 
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]

Reply via email to