Apache Lucene indexing exhibits problems when storing to CNS/GlusterFS. This even leads to ES not running in some scenarios: https://bugzilla.redhat.com/show_bug.cgi?id=1430659. Single node Lucene does work, but is accompanied with errors and delays. NFS is faster, but still slows down indexing noticeably.

Any reason why indexing can't be ephemeral? In the case of a single node, we can store indexing data within the container and re-index if it is restarted. Perhaps this should also be done for ES?