This initialization action installs H2O Sparkling Water on all nodes of Google Cloud Dataproc cluster. This initialization works with Dataproc image version 1.3 and newer.

⚠️ NOTICE: See best practices of using initialization actions in production.

You can use this initialization action to create a new Dataproc cluster with H2O Sparkling Water installed by:

1. Use the gcloud command to create a new cluster with this initialization action.

   To create Dataproc 1.3 cluster and older use conda initialization action:

   REGION=<region>
   CLUSTER_NAME=<cluster_name>
   gcloud dataproc clusters create ${CLUSTER_NAME} \
       --image-version 1.3 \
       --metadata 'H2O_SPARKLING_WATER_VERSION=3.28.0.1-1' \
       --scopes "cloud-platform" \
       --initialization-actions "gs://goog-dataproc-initialization-actions-${REGION}/conda/bootstrap-conda.sh,gs://goog-dataproc-initialization-actions-${REGION}/h2o/h2o.sh"

   To create Dataproc 1.4 cluster and newer use ANACONDA optional component:

   REGION=<region>
   CLUSTER_NAME=<cluster_name>
   gcloud dataproc clusters create ${CLUSTER_NAME} \
       --image-version 1.4 \
       --optional-components ANACONDA \
       --metadata 'H2O_SPARKLING_WATER_VERSION=3.28.0.3-1' \
       --scopes "cloud-platform" \
       --initialization-actions "gs://goog-dataproc-initialization-actions-${REGION}/h2o/h2o.sh"

2. Submit sample job.

   REGION=<region>
   CLUSTER_NAME=<cluster_name>
   gcloud dataproc jobs submit pyspark \
       --cluster ${CLUSTER_NAME} \
       "gs://goog-dataproc-initialization-actions-${REGION}/h2o/sample-script.py"