Tuning Datarobot Clusters¶

Configmaps¶

Datarobot primarily uses configmaps to populate environment variables in the cluster pods. Those variables are then read by services to determine their settings.

There are a lot of configmaps setup and used by Datarobot helm charts. The cluster administrator is able to configure or tune the services using the following configmaps:

• datarobot-modeling-envvars

To add the desired changes to the configmap a cluster admin will need to provide the proper values in a yaml file, and apply a helm upgrade to the release used.

datarobot-modeling-envvars¶

The datarobot-modeling-envvars config map is controlled in umbrella charts with the core.config_env_vars settings. When helm processes the release and its charts, the settings will be put into the datarobot-modeling-envvars.data section of the configmap.

This is very similar to the way the app_configuration.drenv_override section(s) of the config.yaml worked in Datarobot releases prior to 9.0.

To add a configuration option, create or add the core.config_env_vars setting in the values.yaml used to apply the relevant umbrella chart. In the example below the chart is the datarobot-generic chart, and minio is used for object storage.

core:
  config_env_vars:
    S3_HOST: core-minio.DR_CORE_NAMESPACE.svc.cluster.local
    LOGGING_LEVEL: INFO
    LOG_LEVEL: INFO
    AWS_ACCESS_KEY_ID: miniodatarobot
    AWS_SECRET_ACCESS_KEY: miniodatarobot

Say we want to adjust the number of Model Management workers. The default for the chart is two (2) workers. To override that an increase it to four (4) workers, add the MMM_MODMON_WORKER_COUNT key and the 4 value. That will look like this:

core:
  config_env_vars:
    S3_HOST: core-minio.DR_CORE_NAMESPACE.svc.cluster.local
    LOGGING_LEVEL: INFO
    LOG_LEVEL: INFO
    AWS_ACCESS_KEY_ID: miniodatarobot
    AWS_SECRET_ACCESS_KEY: miniodatarobot
    MMM_MODMON_WORKER_COUNT: "4"

To apply the change to the cluster, run the helm upgrade command to upgrade the release.