Pooled storage vs individual mountpoints


I'm setting up a new ES cluster on a set of RHEL servers, each of which 5 x 1Tb disks. I'm trying to work out the best way to configure the storage. Our RHEL admin has described two main options:

  1. Mount each of the 5 disks separately in RHEL, and then list these as 5 separate path.data locations in ES config

  2. In RHEL, configure the 5 disks as a single 5Tb storage pool with a single mount point, and then list this as single path.data location in ES

I can see potential advantages to both. In #1, ES has direct visibility of the underlying disks so this might potentially have performance advantages? However #2 is easier to manage and the storage pool can be expanded or shrunk more easier without ES configuration.

Haven't been able to find much documentation on best approaches with multiple disks per server - appreciate any pointers or comments! Thanks.



Using path.data allows you to stripe your index across multiple disks.

RAID0 will be slightly more performant
path.data will allow a node to continue to function in case one of the local disk fails.

With path.data if one of the 5 disks fails you will have basically only 20% of the data to copy again. With RAID0 you will most likely have to copy 100% of the data.

I'd recommend using path.data.



This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.