Create instance configurations

In the Input section, construct a query that filters on specific allocator tags.

Pick the products and features of the Elastic Stack that can get deployed on the allocators you identified in the previous step. For products such as Elasticsearch, you can also select some additional options, such as the specific node types that can be deployed.

Note that not all combinations of Elasticsearch node types are available. You can create either a general purpose Elasticsearch node that includes all three of data, master, and coordinating, or a dedicated node that includes any one of these types. Machine learning is also available as a separate instance type.

Adjust how memory and storage resources get sized relative to each other and set the available sizes, including the default size. Size your instance configuration so that it will use the available memory and storage on your allocators efficiently, without leaving hardware resources unused. Keep in mind that very small sizes might not provide adequate performance for some use cases.

The size of an instance configuration also determines performance, as CPU resources get sized in lockstep. For example: A 32 GB instance configuration receives double the CPU resources of a 16 GB one.

Obtain the existing instance configurations to get some examples of what the required JSON looks like. You can take the JSON for one of the existing configurations and modify it to create a new instance configuration, similar to what gets shown in the next step.

curl -k -X GET -H "Authorization: ApiKey $ECE_API_KEY" https://COORDINATOR_HOST:12443/api/v1/platform/configuration/instances

Post the JSON for your new instance configuration.

The following examples creates an instance configuration for machine learning with size increments that start at the recommended minimum of 16 GB of memory. To make sure that machine learning nodes get deployed only on the right allocators, this instance configuration also filters for allocator tags from our earlier example to match only allocators with high CPU resources and SSD storage.

curl -k -X POST -H "Authorization: ApiKey $ECE_API_KEY" https://COORDINATOR_HOST:12443/api/v1/platform/configuration/instances -H 'content-type: application/json' -d '{
 "name": "Machine Learning Only",
  "description": "Custom machine learning instance configuration",
  "storage_multiplier": 32.0,
  "discrete_sizes": {
    "sizes": [16384, 32768, 65536],
    "default_size": 16384,
    "resource": "memory"
  },
  "allocator_filter": {
    "bool": {
      "must": [{
        "bool": {
          "must": [{
            "nested": {
              "query": {
                "bool": {
                  "must": [{
                    "term": {
                      "metadata.key": {
                        "value": "SSD"
                      }
                    }
                  }, {
                    "term": {
                      "metadata.value.keyword": {
                        "value": "true"
                      }
                    }
                  }]
                }
              },
              "path": "metadata"
            }
          }]
        }
      }, {
        "bool": {
          "must": [{
            "nested": {
              "query": {
                "bool": {
                  "must": [{
                    "term": {
                      "metadata.key": {
                        "value": "highCPU"
                      }
                    }
                  }, {
                    "term": {
                      "metadata.value.keyword": {
                        "value": "true"
                      }
                    }
                  }]
                }
              },
              "path": "metadata"
            }
          }]
        }
      }]
    }
  },
  "node_types": ["ml"], 
  "instance_type": "elasticsearch"
}'

After you have created your new instance configuration, you can use it when you create new deployment templates or when you edit existing ones.