Basic operations

This guide provides an introduction to creating, scaling, and using Gradient Deployments.

Overview

Gradient Deployments can be created in the Gradient console or via the Gradient CLI.

The default YAML spec deployment available in the Gradient console.

How to create a deployment in the Gradient console

How to create a deployment using the Gradient CLI

You can create deployments using the Gradient CLI and linking an API key. You will also need an existing Project to run your deployment.

1. Create a YAML spec to manage your deployment

image: lucone83/streamlit-nginx
port: 8080
env:
  - name: ENV
    value: VAR
resources:
  replicas: 1
  instanceType: C4

2. Start a deployment

gradient deployments create --name [deployment-name] --projectId [project-id] --spec deployment.yaml

3. View your running deployments and logs

You can view your running deployment in the console UI, which will information on the current spec applied and the current status of your deployment.

You can also see the status of your deployment via the CLI

gradient deployments get --id [deployment-id]

{
    "id": "6771ef0c-147d-422b-8ff3-fef2809b6b26",
    "name": "starter-deployment",
    "deploymentSpecs": [
        {
            "id": "02fe5075-ac1e-4d69-b9c9-4fccc2e75776",
            "data": {
                "image": "lucone83/streamlit-nginx",
                "port": 8080,
                "resources": {
                    "instanceType": "C4",
                    "replicas": 1
                },
                "command": null,
                "env": [
                    {
                        "name": "ENV",
                        "value": "VAR"
                    }
                ],
                "models": null
            },
            "endpointUrl": "d6771ef0c147d422b8ff3fef2809b6b26.clg07azjl.paperspacegradient.com",
            "actor": {
                "avatarUrl": "https://s3.amazonaws.com/ps.profile.images/profileImages/production/te7molbs/2020-08-03T15%3A09%3A11.236Z.png",
                "fullName": null
            },
            "cluster": {
                "id": "clg07azjl"
            },
            "deploymentRuns": [
                {
                    "id": "19f5fec4-142c-4e09-a6ce-9800ffe0d9ca",
                    "availableReplicas": 1,
                    "readyReplicas": 1,
                    "replicas": 1,
                    "deploymentRunInstances": [
                        {
                            "id": "0831f045-a876-4b99-b60c-ce2b653b2825",
                            "phase": "Running",
                            "dtStarted": "2021-09-21T17:47:36.000Z",
                            "dtFinished": null
                        }
                    ]
                }
            ]
        }
    ]
}

How to scale a deployment

You can manage the state of your deployment by updating your spec file and updating the deployment. To stop your deployment from running you can set the replicas to 0, which turns off all workloads. You are able to edit your deployment spec from the Web UI, or using the Gradient CLI

Web UI

You can click the edit button in the Web UI to update your deployment spec to change the number of desired replicas as well as the other parts of the spec.

Gradient CLI

gradient deployments update --id [deployment-id] --spec update-deployment.yaml

How to use a deployment within Gradient Workflows

You can manage deployments in a workflow using the Gradient CLI. Your Paperspace API key and Project ID are available as environment variables

jobs:
  createdDeployment:
    uses: script@v1
    with:
      script: |
        cat > ./deployment.yaml <<EOF
        image: paperspace/first-order-model:server
        port: 8000
        resources:
          replicas: 1
          instanceType: P5000
        EOF
        gradient deployments create --name my-deployment --projectId ${PROJECT_ID} --spec ./deployment.yaml
      image: paperspace/gradient-sdk