Hi everyone, Recently we are implementing Apache Storm Topologies using our own POJOS (and DTOs) as a tuples between BOLTS.
So, until know we are not having issues with serialization (Storm internal, or not detected), we are testin that in our preproduction cluster (1 nimbus and 2 supervisors) after local mode. But a the Question is: If you use your own POJO (Complex or simple) is always necesary to write your Serializer for each, I mean, Kyro or others?. Why I undertand, Apache Storm use Kyro only whe your tuple between bolts and Spouts are primitive types, strings, byte arrays, ArrayList, HashMap, HashSet, and the Clojure collection types ( http://storm.apache.org/releases/1.0.3/Serialization.html). So, if you pass your own POJO between Spouts and Bolts it try to serialize using Java Serizalization because I tell it when I implement my POJOs ... public class MyPOJO implements Serializable { ... Right? Regards, Christian
