Hi Lasse, You may use define a type information factory with config [1] starting from v1.20. We are also working on extending the Flink type system to cover basic generic collection types and avoid falling back to Kryo.
[1] https://nightlies.apache.org/flink/flink-docs-master/docs/dev/datastream/fault-tolerance/serialization/types_serialization/#defining-type-information-using-a-factory Best, Zhanghao Chen ________________________________ From: Lasse Nedergaard <lassenedergaardfl...@gmail.com> Sent: Friday, September 13, 2024 14:27 To: user <user@flink.apache.org> Subject: Recommendations for avoid Kryo Hi. I was wondering how others handle the situation where you flink job has to read or write a data structure in json mapped to a pojo defined in an external library and these external pojo’s contains data types, like uuid, generic list etc, that you need to annotate with type information in order to avoid Kryo serialisation. Is best practice to reimplement the pojo’s in the flink job with the type annotation? This solution works but then you have the definition twice and when changes occur you have to remember to change it both places. Or do we have other options. What is the recommended what of handling this problem? Med venlig hilsen / Best regards Lasse Nedergaard
