View deployments

This page explains how to view and manage deployments in H2O MLOps. It covers how to access deployment details, monitor performance, perform scoring, and carry out common tasks such as downloading logs or deleting deployments.

To view the list of available deployments, click Real-time deployments on the left navigation menu.

Understand the Deployments page

To view details of a specific deployment, click the name of the deployment you want to view.

The Deployment details page is divided into the following three tabs:

• Details
• Endpoints
• Quick scoring

Details

The Details tab provides key information about the deployment, including configuration, status, and advanced settings.

• Deployment name: The name of to the deployment.

• Deployment description: A brief summary of the deployment's purpose or functionality.

• Deployment type: The deployment type for the deployment (Single model, A/B test, or champion/challenger).

• Status: The current status of the deployment. For more information, see States.

• Deployed model details: Information about the model associated with the deployment:

  • Model name: The name of the deployed model.
  • Version: The specific version of the model in use.
  • Artifact type and runtime: The artifact type and runtime for the deployment.

• Advanced settings: Additional configuration options for the deployment. You can update the following:

  • Kubernetes options
    • Replicas: The number of static replicas of the deployed model.
    • Requests: The amount of resources the deployment requires from Kubernetes.
    • Limits: The maximum amount of resources that are available to the deployment.
  • Model monitoring: To enable or disable model monitoring, use the Enable Monitoring toggle. When enabled, scoring data is collected and stored.
  • Security
    • Level: The security level specified when the deployment was created.
    • Passphrase: If you selected the Passphrase (Stored as plain text) option when creating the deployment, the passphrase can be viewed in the Security Details section. Note that if you select the Passphrase (Stored hashed) option, the passphrase cannot be viewed.

  After updating the advanced settings, click Save changes.
  For more information on the advanced settings, see Advanced settings.

Download deployment logs

To download deployment logs, click Download Logs in the Details tab. The logs are downloaded to your device as a ZIP file.

Delete a deployment

To delete a deployment, click the Delete button in the Details tab.

States

The following is a list of possible states for deployments in MLOps:

• Preparing - The deployment is being prepared for launch
• Launching - The deployment is launching to an environment
• Failed - The deployment failed during preparation or launch
• Healthy - The deployment is alive and healthy
• Unhealthy - Health issues have been detected with the launched deployment
• Terminating - The deployment is terminating
• Pending - The deployment has been created and is awaiting processing

Endpoints

From the Endpoints tab, you can view and copy the default deployment URL. If the default scorer endpoint is not suitable, you can add a custom endpoint with a different path.

Add custom endpoint

1. Click Add custom endpoint.
2. Enter a name for the new endpoint.
3. (Optional) Add a description for the endpoint.
4. Specify the endpoint path.
5. Click Create endpoint.

Scoring with endpoint URL

In the Endpoints tab, copy the endpoint URL.

You can use the URL in your own application for sending scoring requests.

Quick scoring

Scoring with cURL scoring request

The cURL scoring request is auto-generated based on the scoring configuration.

1. In the Quick scoring tab, copy the sample curl scoring request.
2. Open a Terminal window and paste the sample curl request to view the model scoring information.

Score directly from the UI

To score the deployment directly from the UI, click Score in the Quick scoring tab.

After scoring, the results are displayed in the interface.

Missing values

To indicate a field as a missing value for any of the following runtime options, use the corresponding value displayed in the following table.

note

The information provided in this section is only guaranteed to work if no changes have been made to the default missing_values configuration setting in Driverless AI (DAI). If you have changed the missing_values configuration setting in DAI, contact the H2O support team for assistance.

Runtime	Numeric missing value	String missing value
Driverless AI MOJO scorer	Empty string	Empty string
Python Pipeline scorer (MLOps 0.57.3 and earlier)	"1.7976931348623157e+308"	Empty string
Python Pipeline scorer (MLOps 0.58.0 and later)	"1.7976931348623157e+308" or empty string	Empty string
H2O-3 MOJO	Empty string	Empty string
MLflow/ .pkl file	Not supported	Empty string

---

Feedback

• Submit and view feedback for this page
• Send feedback about H2O MLOps to cloud-feedback@h2o.ai