This is a Terraform module that deploys H2O.ai Driverless AI on Oracle Cloud Infrastructure (OCI). It is developed jointly by Oracle and H2O.ai. For instructions on how to use this material and details on getting support from the vendor that maintains this material, please contact them directly.
Instructions on how to use it are below. In addition, here's a video walkthrough:
First off you'll need to do some pre deploy setup. That's all detailed here.
DAI requires a license key. If you don't already have a key, you can get a trial key. You can deploy these templates before you get a key, but you'll be prompted for a key at first login.
Now, you'll want a local copy of this repo. You can make that with the commands:
git clone https://github.com/oracle/oci-quickstart-h2o.git
cd oci-quickstart-h2o/terraform
ls
That should give you this:
We now need to initialize the directory with the module in it. This makes the module aware of the OCI provider. You can do this by running:
terraform init
This gives the following output:
Now for the main attraction. Let's make sure the plan looks good:
terraform plan
That gives:
The next command will deploy H2O DAI using the values in the variables.tf file:
key: Set to the value of your key. If empty, you will be prompted to enter your key at first login.shape: DefaultBM.GPU2.2, these templates support both CPU and GPU (higher performance) shapes.ad_number: Default0, choose a value where you have quota for the desired shape.disk_size_gb: Default0, size of block volume in GB for data, min 50. If set to 0 volume will not be created/mounteduser: Defaultadmin, these templates set up local authentication using file/etc/dai/htpasswdpassword: Defaultadmin
If that's good, we can go ahead and apply the deploy:
terraform apply
You'll need to enter yes when prompted. The apply should take about seven minutes to run. Once complete, you'll see something like this:
When the apply is complete, the infrastructure will be deployed, but the cloud-init scripts will still be running. Those will wrap up asynchronously. So, it'll be a few more minutes before your DAI instance is accessible. Now is a good time to get a coffee.
The apply prints the URL of H2O Driverless AI when it completes. The deployment creates a self-signed certificate, so you'll need to confirm the exception.
Logging into that we see:
After accepting the agreement, we can see the login screen. Use the login defined in the variables.tf file:
The documentation here provides a walkthrough and some good next steps. There are a bunch of public data sets to get you started. Here's an example of an experiment I set up on one using that walkthrough.
This machine is using Oracle Linux (OL). The default login is opc. You can SSH into the machine with a command like this:
ssh -i ~/.ssh/oci opc@<Public IP Address>
H2O is installed under /opt/h2o. Logs from the install are under /var/log/messages and viewable with root privileges.
You can also login to the web console here to view the IaaS that is running DAI.
When you no longer need the deployment, you can run this command to destroy it:
terraform destroy
You'll need to enter yes when prompted. Once complete, you'll see something like this:
