Consider a cache composed of two entities A and B, entity A is stored on RAM and entity B on Infinispan itself. The configuration to achieve that is below:

cacheCfg.indexing()
    .index(Index.ALL)
    .addIndexedEntity(A.class)
    .addIndexedEntity(B.class)
    .addProperty("A.directory_provider", "ram")
    .addProperty("A.indexmanager", "org.infinispan.query.indexmanager.InfinispanIndexManager")

The configuration Index.ALL only makes sense to RAM directory, while for InfinispanIndexManager it'd cause redundant indexing. The indexing mode should not be at cache level but at "index" level