flambo

sorenmacbeth 2016-09-01T01:07:58.000009Z

@turbopape 👍

2016-09-01T10:38:02.000010Z

Hi guys, I have a problem

2016-09-01T10:38:23.000011Z

When using mllib, I run into incmpatibilities with kryo serialization

2016-09-01T10:38:43.000012Z

Seems it’s a known bug, and I need to be going with spark 2.X

2016-09-01T10:39:11.000013Z

problem is, When I reaquire spark 2.X, it tells me that it can’t find logger class...

2016-09-01T10:39:14.000014Z

any idea ?

2016-09-01T10:43:49.000015Z

here is the relevant trace:

2016-09-01T10:43:51.000016Z

com.esotericsoftware.kryo.KryoException: java.lang.IllegalArgumentException: Can not set final scala.collection.mutable.ListBuffer field org.apache.spark.mllib.fpm.FPTree$Summary.nodes to scala.collection.mutable.ArrayBuffer Serialization trace: nodes (org.apache.spark.mllib.fpm.FPTree$Summary) org$apache$spark$mllib$fpm$FPTree$$summaries (org.apache.spark.mllib.fpm.FPTree) at com.esotericsoftware.kryo.serializers.FieldSerializer$Obje

2016-09-01T10:44:02.000017Z

seems it does not like fptree

2016-09-01T10:52:45.000018Z

seems to have been better if I register the missing classes at kryo’s level :

2016-09-01T10:53:03.000019Z

… (conf/set "spark.kryo.classesToRegister" "scala.collection.mutable.ArrayBuffer,scala.collection.mutable.ListBuffer")

2016-09-01T10:53:39.000020Z

But still waiting for it to finish ...

2016-09-01T11:01:34.000021Z

Yes.

2016-09-01T11:01:40.000022Z

This definitely worked !