How to reproduce

Environment setup

Clone ignite and compile

Download spark 2.4.3 without Hadoop

Download hadoop 3.1.2

Install sbt 1.2.8

Compile and create a fat jar

sbt assembly

Lauch ignite

$IGNITE_HOME/bin/ignite.sh

This will run successfully

java -cp target/scala-2.11/ignite-failed-unmarshal-discovery-data-assembly-0.1.0-SNAPSHOT.jar Cacher

This will fail unexpectedly

$SPARK_HOME/bin/spark-submit target/scala-2.11/ignite-failed-unmarshal-discovery-data-assembly-0.1.0-SNAPSHOT.jar

See also

lauch.sh to get few more details on setup