ML is available but not enabled

[robertvolkmann]

Hi everyone,

we started with our trial licence to evaluate the machine learning features. But after following the elasticsearch documentation machine-learning remains disabled.

Start point:
Elasicsearch 6.6.2
xpack.ml disabled on all nodes
node.ml disabled on all nodes

We do the following steps:

1. We enabled xpack.ml on all master nodes (node.master: true) and we do a rolling restart of them
2. We enabled xpack.ml and node.ml for exactly one node
3. We enabled xpack.ml for kibana

Outcome:
_xpack:

...
"ml" : {
  "description" : "Machine Learning for the Elastic Stack",
  "available" : true,
  "enabled" : false,
  "native_code_info" : {
    "version" : "N/A",
    "build_hash" : "N/A"
  }
},
...

_xpack/usage

...
  "ml" : {
"available" : true,
"enabled" : true,
"jobs" : {
  "_all" : {
    "count" : 0,
    "detectors" : {
      "total" : 0.0,
      "min" : 0.0,
      "avg" : 0.0,
      "max" : 0.0
    },
    "model_size" : {
      "total" : 0.0,
      "min" : 0.0,
      "avg" : 0.0,
      "max" : 0.0
    },
    "forecasts" : {
      "forecasted_jobs" : 0,
      "total" : 0
    }
  }
},
"datafeeds" : {
  "_all" : {
    "count" : 0
  }
},
"node_count" : 1

},
....

Should we enable xpack.ml on all data nodes? (node.master: false, node.data: true)

[richcollier]

See: https://www.elastic.co/guide/en/elasticsearch/reference/current/ml-settings.html

The setting that you're probably not setting correctly is:

xpack.ml.enabled: true

It is not called "xpack.ml"

[robertvolkmann]

Thank you for your suggestion, but it is already in place and does not work. Below are the truncated responses of the cluster configuration.

GET _cat/nodes

10.81.13.41 23 100  6 0.12 0.18 0.14 di - db6
10.81.13.36 39  99 11 0.41 0.40 0.41 m  - db1-master
10.81.13.36 41  99 13 0.41 0.40 0.41 di - db1
10.81.13.38 10  99  9 0.21 0.15 0.14 di - db3
10.81.13.39 12  99 10 0.24 0.34 0.43 di - db4
10.81.13.37 26  99  6 0.04 0.10 0.11 di - db2
10.81.13.37 45  99  6 0.04 0.10 0.11 m  - db2-master
10.81.13.38 28  99  9 0.21 0.15 0.14 m  * db3-master
10.81.13.40 37  41  0 0.00 0.01 0.05 i  - db5

GET _nodes/db1-master/settings
GET _nodes/db2-master/settings
GET _nodes/db3-master/settings
GET _nodes/db5/settings

"xpack" : {
  "watcher" : {
    "enabled" : "false"
  },
  "security" : {
    "enabled" : "false"
  },
  "monitoring" : {
    "enabled" : "true",
    "collection" : {
      "enabled" : "true"
    }
  },
  "ml" : {
    "enabled" : "true"
  }
},

GET _nodes/db1/settings
GET _nodes/db2/settings
GET _nodes/db3/settings
GET _nodes/db4/settings
GET _nodes/db6/settings

"xpack" : {
  "watcher" : {
    "enabled" : "false"
  },
  "security" : {
    "enabled" : "false"
  },
  "monitoring" : {
    "enabled" : "true",
    "collection" : {
      "enabled" : "true"
    }
  },
  "ml" : {
    "enabled" : "false"
  }
},

[richcollier]

Please refer to:

Separate machine learning node configs

Hello i have a 3 node cluster up and running and on all the nodes node.ml is set to "false".Now i want to include a extra node into cluster which will only perform machine learning jobs. As far as my understanding node.ml is to make the node as machine learning node and xpack.ml.enabled is to make the node to perform machine learning jobs. SO setting the node.ml true for one node which is the new node i am introducing and xpack.ml.enabled true on all nodes should make the cluster with one ml no…

In that post you should see that:

xpack.ml.enable should be true on all nodes as it is required for the nodes to talk 'ml' to each other. node.ml should only be set on the node you want the machine learning jobs to run on.

Seems like you should set xpack.ml.enable to be true on all nodes and then only set node.ml to be true on db5

[robertvolkmann]

Thank you for your suggestion. Yesterday evening I activate xpack.ml.enable on db1 and nothing changed. But waiting ten hours(*) changed everything. It works.

(*) Maybe our cluster is an old man.

Greetings