Dedicated master - quorum and recovery settings

Elastic Stack Elasticsearch
1

My cluster has finally reached the size/complexity where I'm ready to roll out dedicated master nodes.

What I haven't been able to 100% nail down is what settings belong on the data nodes vs. what belong on the master nodes.

Here are the settings in question:

gateway.expected_master_nodes
gateway.recover_after_nodes
gateway.expected_nodes
gateway.recover_after_time

My question is simply - which of these settings need to exist on the data nodes, master nodes, or both?

My assumption would be:

Data nodes only
gateway.recover_after_nodes
gateway.expected_nodes
gateway.recover_after_time

Master nodes only
gateway.expected_master_nodes

While I've found a decent amount of material encouraging the use of dedicated masters, the documentation for how to pull that off (correctly) is a bit lacking.

Thanks in advance!

2

You need the same settings on all the nodes :smile:

Fair point around the docs, will see what we can do to fix it!

3

Thanks. Every node it is.

I've seen a lot of improvements in the docs over the past year or so, I think a few more "here's what configuration settings apply to what nodes in the scenarios we've been describing" would go a long way.

4

Understood, problem is that there is a lot of different scenarios.
However when it comes to basic N node deployments this should be easier to do.