New Cloudera Data Visualization (CDV) Runtime releases are automatically added to your
deployment when an internet connection is available. However, in air-gapped Cloudera Data
Science Workbench (CDSW) deployments, if you want to use CDV, you need to manually load the CDV
Visualization Runtime in the cluster.
Ensure you have the following:
- CDSW 1.9.0 or higher to support runtimes
- Root access to all cluster nodes
- CDSW installed as a parcel
- Admin access on the CDSW cluster
- Proficiency in Docker, SQL, and Linux
-
Download the
repo-assembly.json
file from the ‘artifacts’ directory of the latest CDV
version.
-
Download the Docker image to an Internet-connected node or your local machine.
image_identifier=$(jq -r '.runtimedataviz[0].image_identifier' repo-assembly.json)
docker pull "${image_identifier}"
-
Load the Docker image to all cluster nodes (master and all workers) using the
docker save and docker load commands.
docker save -o runtimedataviz.tar <image_name>
docker load -i runtimedataviz:image_identifier
-
Verify that the Docker image is available on all nodes and has its original name and
tag using the docker images command.
You will get a summary view of the Docker images, showing details such as the
repository name, tag, image ID, creation date, and size.
-
Add the Cloudera Data Visualization image as a custom runtime, using the original
Docker image name.
For example:
docker.repository.cloudera.com/cloudera/cdv/runtimedataviz:7.1.2-b53
For detailed instructions, see Adding New ML Runtimes.