Machine learning nodes

[mkemalm]

Hello,

i am asking to verify some considerations about machine learning in elk stack.

-xpack.ml.enabled should be true in master nodes. True or False?
-node.ml should be true in nodes which are doing ml job actually. True or False?
-is it better to use dedicated nodes for node.ml or using data nodes for it? Is it better to have multiple of them?

Regards.

[JKhondhu (Jymit Singh Khondhu)]

A machine learning node is a node that has xpack.ml.enabled and node.ml set to true, which is the default behavior. If you set node.ml to false, the node can service API requests but it cannot run jobs. Taken from: https://www.elastic.co/guide/en/x-pack/current/ml-concepts.html

0. False
1. True
2. Yes to the use of dedicated nodes if:

ML can be setup as it's own dedicated node type, similar to what you can do now for master or data nodes. So, you will have a few options depending on how your cluster is currently running. For example:

• If the current node(s) do not have a lot of idle CPU, then:
  To run ML on the same data node, you could get machines with more cores as well as increase the RAM
  or run ML as a dedicated node in your cluster.
• if the current node(s) do have a lot of idle CPU, then:
  you could simply increase the memory alone.
  A lot of this will depend on how many 'jobs' you configure, what they are configured to do given the source data set and how often they will run. The best way to start will be to assess what types of jobs you will plan on running to determine what the suitable configuration will be when ML is deployed.