You shouldn't use file backups for Elasticsearch, but rather the snapshot and restore APIs. This can be done by way of other tools, like Elasticsearch Curator.
The primary reason to not use a file-type backup approach is that the data would very likely be corrupted. The Lucene data structures in the should-never-be-touched data paths are in constant change so long as indexing is going on. If one file were backed up while another were changing, then there would be a mismatch, and corruption would ensue. Your file-based backup would be worthless.