Shared Data-Store for multiple Nodes

Hi All,
At the moment we have 3 Elasticsearch-Node (All Master)

Each Elasticsearch-Node is installed on a Ubuntu-VM on our central ESXi-Cluster.
Each node has a dedicated /data mount to store the Data on a own volume in our SAN.
The current default for our indices is 1 Replica.
So i thought about one Central data-store for all nodes, so i could reduce the storage Size by the half because each shard is not stored 2x on different data-volumes.

I found the setting for that in the documentation:

So its possible and (at the moment) supported.
But did "This setting is deprecated in 7.x and will be removed in version 8.0" mean that in 8.0 it wouldn't be supported anymore or just you don't need to set this attribute?

And why its not recommend by Elastic, what are possible downsides?

Thanks in Advice,

Your setup does not have that effect. If you have 1 replica (and 1 primary, obviously) then every shard is stored twice regardless of the underlying storage layout.

It won't be supported any more. Each node will require its own dedicated data path.

If using max_local_storage_nodes then the relationship between nodes and their data depends on the order in which they start up and is therefore not well-defined. It's all too easy to end up with a cluster that cannot recover after a restart.

1 Like

Hi David,
Thanks for your reply.
I forgot to write that I thought to reduce to 0 replicas because every node would have all data to use if one node goes offline.

But if it wouldn't be supported anymore we won't change that😅

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