Gravity Installation#

Gravity is an open-source, containerized Kubernetes framework that allows containerized applications like Anaconda Enterprised to be delivered in a self-contained installer that includes a special containerized implementation of Kubernetes. Anaconda has shipped Data Science & AI Workbench in a Gravity package since its first release.

Due to changes in the business model of Teleport, the original developers of Gravity, we are limiting the use of the Gravity-based installer to existing customers. For more information, see Anaconda’s Gravity update policy.

New customers, and any existing customers who no longer want to use Gravity, are encouraged to explore our Bring Your Own Kubernetes (BYOK8s) option.

After you have verified that your environment is prepared for installation by completing the pre-installation checklist, you’re ready to install the cluster!

Note

The Gravity installation process only sets up a single control plane node. Once installation is complete, you can add additional nodes as desired, including GPUs.

Open a terminal on the master node and log in to a service account with sudo access.
Download the installer by running the following command:
```
# Replace <INSTALLER_LOCATION> with the provided location of the installer
curl -O <INSTALLER_LOCATION>
```
Note

If you are in an airgapped environment, you must deliver the installer to the master node in an alternate manner.

Decompress the installer and enter the top directory by running the following commands:

# Replace <INSTALLER> with the installer file you just downloaded
tar xzvf <INSTALLER>
cd <INSTALLER>

Set up and initialize your conda environment by running the following command:
```
./conda-bootstrap.sh
```
Enter yes to allow the installer to initialize your ae5-conda environment by running conda init when prompted.
Verify your environment successfully created by running the following command:
```
conda env list
```
Note

This command returns a list of existing conda environments. You should see a base environment present with an asterisk next to the directory to indicate it is currently active.
Perform the Anaconda preflight system check on the master node by running the following command:
```
ae-preflight
```
Note

You will receive an overall PASS or WARN result from the script. The output is captured in a results.txt file for later reference, if necessary. If something goes wrong, you can view the results to determine the issue and correct it.
Create a file named values.yaml and populate it with the following content:
```
# Replace <HOSTNAME> with the fully qualified domain name of the host server
# Replace <GID> with the group ID with read-write access to the shared network file system (NFS) volume
# Replace <SERVER_FQDN> with the address of the shared NFS volume
# Replace <SERVER_PATH> with the path to the shared NFS volume
apiVersion: v1
kind: ConfigMap
metadata:
  name: anaconda-enterprise-install
data:
  values: |
    hostname: <HOSTNAME>
    persistence:
      groupID: <GID>
      nfs:
        server: <SERVER_FQDN>
        path: <SERVER_PATH>
```
If you are installing a single-node cluster and you want to use the /opt/anaconda/storage volume for managed persistence, the persistence: section of the values.yaml can be entered as:
persistence: pvc: anaconda-storage
Note

The direct use of /opt/anaconda/storage for managed persistence works only for single-node clusters. It must be replaced with a network file system (NFS) configuration before adding additional nodes to the cluster. Otherwise, sessions and deployments launched on the additional nodes will fail to start. For more information, see managed persistence.

Run the gravity installation command:

# Replace <ADVERTISE-ADDR> with the IP address you want to be visible to the other nodes. If you have a private network and all nodes can communicate on it, use the private IP address for the advertise address
# Replace <CLUSTERNAME> with the name you're giving to your cluster. Alphanumeric characters and periods only
# Replace <VALUES_PATH> with the filepath to the values.yaml file you just created
sudo ./gravity install --advertise-addr=<ADVERTISE-ADDR> --cluster=<CLUSTERNAME> --config <VALUES_PATH> --service-uid=$(id -u) --cloud-provider=generic

Here is an example of what the installation command looks like for a single-node cluster:

sudo ./gravity install --advertise-addr=192.168.1.1 --cluster=MyCluster --config ./values.yaml --service-uid=$(id -u) --cloud-provider=generic

If you are using an alternate TMPDIR, prepend the command with this new value. For example:

sudo TMPDIR=/mytmp ./gravity install --advertise-addr=192.168.1.1 --cluster=MyCluster --config=./values.yaml --service-uid=$(id -u) --cloud-provider=generic

Caution

The --service-uid is the service account UID and cannot be a root user. If you are performing a reinstallation of Workbench, you must delete the previous planet user prior to setting the new --service-uid.

Monitor the progress from the command line. Your output will look like this:

* [0/100] starting installer
* [0/100] preparing for installation... please wait
* [0/100] application: AnacondaEnterprise:5.5.x
* [0/100] starting non-interactive install
* [0/100] initializing the operation
* [20/100] configuring packages
* [50/100] installing software

Allow time for the process to complete (approximately 20-30 minutes).
Create a local user account and password to log in to the Gravity Operations Center. To do so, enter the Gravity environment from any of the nodes by running the following command:
```
sudo gravity enter
```

Create a local user account and password by running the following command:

# Replace <EMAIL> with the email address for the account
# Replace <PASSWORD> with a password for the account. The password must be six characters long
gravity --insecure user create --type=admin --email=<EMAIL> --password=<PASSWORD> --ops-url=https://gravity-site.kube-system.svc.cluster.local:3009

Inform gravity that you have completed the installation by running the following command from within the Gravity environment:
```
gravity site complete
```
Proceed to the post-install configuration steps.

Note

Workbench itself will likely require another 30 minutes or so to fully load. The post-install configuration steps describe how to confirm that this process is fully complete.