Indexer job via MiddleManager


Is there any configuration that let me run the indexer from MiddleMaanger without isolating the Hadoop Indexer Class Loader from CliPeon Class Loader? or any plans to add one?

I found an issue with MapR Client, I’m trying to use MapR as deep storage and also the Hadoop cluster for the indexer. and I found a limitation in the case that one of the extension is using JNI.

MapR client is using JNI, so when CliPeon start it load HDFS as deep storage and this init MapR client and load MapR native library. When the indexer starts to work it also try to load the extensions in a new Class loader and failed to load the JNI.

When I add HDFS deep storage to the class path (load it under root class loader and did a code change in buildClassLoader to not isolate the class loader it working for me (I also add to change JobHelper to add a case for maprfs scheme)

Does anyone encounter this issue?