Node roles impact on nodes

Hi,

We currently have a really big cluster with 150+ nodes.
We are using node attributes to manage the data tiers and our ILM is based on it (node.attr.data).

We are currently investigating the impact of migrating to "node roles" which are recommended by Elasticsearch. All our data nodes are currently using the "data" role.

If we specify a more precise role like data_content, data_hot... on our nodes, will it improve our cluster performances ?
Is there some internal optimization in Elasticsearch based on the new node roles or is it only an informative data used on ILM (just like node.attr.data)?

Thanks for your time ! :slight_smile:

There is no optimization, it is just informative data to be used in ILM and default templates.

A couple of months ago I did the same migration, went from custom attributes to using the built-in roles, what I did was the following:

  • Change the roles for every node in elastcsearch.yml while keeping the custom attributes.
  • Restart every node, this was done while upgrading the cluster.
  • After all nodes were restarted with new roles, change all the ILM from custom attributes to the default roles.
  • Remove the custom attributes from elasticsearch.yml, but the nodes will only be restarted on the next planned upgrade.

Thank you for your answer :slight_smile:

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