Using NiFi 1.18.0/1.20.0 and trying to run a Groovy Script to access a Hazelcast 3.9 cluster with a composite Java key.
I started with Matt Burgess' link (http://funnifi.blogspot.com/2016/02/executescript-using-modules.html) to try and work through the Hazelcast connection via Groovy. It appears that ExecuteScript picks up the JARs in the Module Directory listed, but I have issues at that point with a HazelcastSerialization issue. The cache composite Java keys are in a JAR file also located in the Module Directory but the error still exists. If I move the Hazelcast 3.9 JARs as well as the composite Java keys to the NiFi ./lib directory, the HazelcastSerialization error doesn't occur. But, from this StackOverflow link (https://stackoverflow.com/questions/45273594/nifi-adding-custom-lib-jar) Matt Burgess says that is not recommended because of class loading situations. I couldn't find anything in the NiFi Admin Guide that talks about a directory to have other JAR files, it only talks about NAR files, so I'm not sure what the "proper" solution would be in this case, outside of building out a NAR with processors to do this work. Also, it doesn't seem possible to connect to a 3.x Hazelcast cluster from a 4.x JAR file. (Maybe I'm mistaken? But config values have changed. Group name -> Cluster name. No Group password, etc.) And we aren't migrating our cache to 4.x yet, so I cannot use the Hazelcast Controller Services. Thoughts on a solution? Thanks, John
