Hello guys,
i want to deploy an Elasticsearch cluster (3 Nodes, ES-Version: 6.x) on Kubernetes and want to know, which storage types are efficient to use for data.
Do you think CephFS would be a proper solution or is it better, to store the data on physical hard disks without a distributed filesystem as abstraction?
Thanks 
It is better to store the data on physical hard disks without a distributed filesystem as abstraction. I don't know of any tests of Elasticsearch on CephFS (for either performance or correctness) but I do know that Elasticsearch is extensively tested against physical disks.
Thx @DavidTurner 
Do you know, if there would be any impact on the performance of the CEPH system itself?
No, I don't know. As I said, I don't know of any tests at all using CephFS with Elasticsearch. You can almost certainly use Rally to perform your own experiments.