Version: 0.4

Configuring Databricks

The following steps enable Tecton to access your data plane account to manage AWS and Databricks resources.

Before you get started:

Decide on a name for your deployment (e.g. mycompany-production). This name will become part of the URL that is used to access your Tecton UI (mycompany-production.tecton.ai). Note: This name must be less than 22 characters.
Determine which AWS region you'd like Tecton deployed into (e.g. us-west-2).

Connecting Tecton to Databricks

When you first log in to your Tecton domain, you'll be directed to the Databricks Cluster Configuration page.

Follow the instructions on the page to give Tecton access to your Databricks workspace. With the SaaS deployment model you'll be able to get started immediately.

Configuring AWS resources

Before launching your first Tecton use-case to production, Tecton requires configuring S3 for your offline feature store. Optionally, you may choose to configure a self-managed online store.

"Terraform Templates for AWS Account Configuration"

If your organization uses Terraform to manage AWS resources, we recommend you leverage this sample terraform setup repository in place of manually entering these values. The instructions below may still be a valuable reference when adapting the template to your needs, especially if your Databricks installation is in a separate account from the Tecton control plane.

At the end of these steps, you will have created:

A S3 Bucket where Tecton will write feature data.
The cross-account role for the Tecton control plane to manage DynamoDB and S3, e.g. tecton-{DEPLOYMENT_NAME}-cross-account-role and tecton-{DEPLOYMENT_NAME}-cross-account-policy.
The Spark policy, e.g. tecton-{DEPLOYMENT_NAME}-spark-policy attached to your existing Databricks Spark role. This policy enables Spark to to write feature data to Dynamo, and read and write feature data to S3.

Create a Tecton S3 Bucket

Tecton will use a single S3 bucket to store all of your offline materialized feature data.

To set this up, create an S3 bucket called tecton-[DEPLOYMENT_NAME] (e.g. tecton-mycompany-production).

Ensure the bucket's region is the same as the region in which you'd like to deploy Tecton (e.g. us-west-2)
Enable default encryption using the Amazon S3 key (SSE-S3)

Create a cross-account role for the Tecton Control Plane

Next, you'll need to create an access policy and an IAM role which the Tecton control plane will use to orchestrate AWS resources in your account.

In the AWS Console of the account you want to deploy Tecton into, go to the IAM service.
Click the Policies tab in the sidebar.
Click Create Policy.
1. Paste in the following JSON policy, replacing ${REGION} with the AWS region you selected for your deployment, ${ACCOUNT} with the account ID of your AWS account, ${DEPLOYMENT_NAME} with your Tecton deployment name, and if using Databricks in the same AWS account, ${SPARK_ROLE} with instance profile role used by Databricks. If your ${SPARK_ROLE} is prefixed with "tecton-" or in a different account, you can omit that line.
  templates/ca_policy.json
  loading...
  See full example on GitHub
2. Click Next: Tags
3. Add the following tag to the policy
```
key: tecton-accessible:DEPLOYMENT_NAME
value: true
```
1. Click Next: Review
2. Give the policy an easy to remember name starting with tecton-, like tecton-{DEPLOYMENT_NAME}-cross-account-policy
3. Click Create Policy
Click the Roles tab in the sidebar.
Click Create role.
1. Under Select type of trusted entity, click the Another AWS account tile.
2. Specify Tecton's Account ID. Please contact your Account Manager to get the appropriate ID.
3. Enable the option "Require external ID."
4. Enter a random External ID of your choice (for example, a UUID works well). Make sure to note down the external ID that you choose -- you'll need to provide this to Tecton to complete the installation.
5. Click the Next: Permissions button
6. Search for the policy you just created (e.g. tecton-{DEPLOYMENT_NAME}-cross-account-policy), and click the check box next to that policy to attach the policy to the new role.
7. Click the Next: Tags button.
8. Click the Next: Review button.
9. In the Role name field, enter a role name starting with tecton-, such as tecton-{DEPLOYMENT_NAME}-cross-account-role.
10. Click Create role. You will see a list of roles displayed.

Create an AWS Policy for Spark

Databricks will need permission to access the S3 and Dynamo resources created by Tecton. To enable this, you'll need to create a policy, and attach that policy to the Spark role Databricks uses for the Spark agents.

In the AWS Console of the account that will be running Spark (this could be different from the Tecton deployment account if you have Databricks deployed in another account), go to the IAM service.
Click the Policies tab in the sidebar.
Click Create Policy.
1. Paste in the following JSON policy, replacing ${REGION} with the AWS region you selected for your deployment, ${ACCOUNT} with the account ID of your Tecton AWS account, and ${DEPLOYMENT_NAME} with your Tecton deployment name
  templates/spark_policy.json
  loading...
  See full example on GitHub
2. Click Next: Tags
3. Add the following tag to the policy
```
key: tecton-accessible:DEPLOYMENT_NAME
value: true
```
1. Click Next: Review
2. Give the policy an easy to remember name starting with tecton-, like tecton-{DEPLOYMENT_NAME}-spark-policy
3. Click Create Policy
4. Attach this policy to the instance profile role used by Databricks.

Cross-Account Databricks Configuration

info

If your Databricks deployment is in a different account to your Tecton deployment you will need to complete this section to create a cross-account role for Databricks to read and write features. Otherwise, you can skip to the next step.

In the AWS Console of the account you want to deploy Tecton into, go to the IAM service.
Click the Policies tab in the sidebar.

Click Create Policy.

Paste in the following JSON policy, replacing ${REGION} with the AWS region you selected for your deployment, ${ACCOUNT} with the account ID of your AWS account, and ${DEPLOYMENT_NAME} with your Tecton deployment name

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "DynamoDB",
      "Effect": "Allow",
      "Action": [
        "dynamodb:ConditionCheckItem",
        "dynamodb:DescribeTable",
        "dynamodb:PutItem"
      ],
      "Resource": [
        "arn:aws:dynamodb:{REGION}:{ACCOUNT}:table/tecton-{DEPLOYMENT_NAME}*"
      ]
    },
    {
      "Sid": "DynamoDBGlobal",
      "Effect": "Allow",
      "Action": ["dynamodb:ListTables"],
      "Resource": "*"
    }
  ]
}

Click Next: Tags
Click Next: Review
Give the policy an easy to remember name starting with tecton-, like tecton-{DEPLOYMENT_NAME}-databricks-cross-account-policy
Click Create Policy

Click the Roles tab in the sidebar.
Click Create role.
1. Under Select type of trusted entity, click the Another AWS account tile.
1. Specify the account id where Databricks is deployed.
2. Enable the option "Require external ID."
3. Enter a random External ID of your choice (for example, a UUID works well). Make sure to note down the external ID that you choose -- you'll need to provide this to Tecton to complete the installation.
4. Click the Next: Permissions button
5. Search for the policy you just created (e.g. tecton-{DEPLOYMENT_NAME}-cross-account-policy), and click the check box next to that policy to attach the policy to the new role.
6. Click the Next: Tags button.
7. Click the Next: Review button.
8. In the Role name field, enter a role name starting with tecton-, such as tecton-{DEPLOYMENT_NAME}-databricks-cross-account-role.
9. Click Create role. You will see a list of roles displayed.
10. Go back to the role's Trust relationships tab, edit trust relationship, and change the Principal from "arn:aws:iam::{ACCOUNT}:root" to "arn:aws:iam::{ACCOUNT}:role/{SPARK_ROLE}"
Add a Policy to the S3 bucket named "tecton-{DEPLOYMENT_NAME}" to allow Databricks to read/write to it.
1. Navigate to S3 -> "tecton-{DEPLOYMENT_NAME}" S3 Bucket -> Permissions. There, add the following policy where {SPARK_ROLE_ARN} refers to the arn of instance profile role used by Databricks
```
{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "DatabricksS3",
      "Effect": "Allow",
      "Principal": {
        "AWS": "{SPARK_ROLE_ARN}"
      },
      "Action": ["s3:GetObject", "s3:PutObject", "s3:DeleteObject"],
      "Resource": "arn:aws:s3:::tecton-{DEPLOYMENT_NAME}/*"
    },
    {
      "Sid": "DatabricksS3List",
      "Effect": "Allow",
      "Principal": {
        "AWS": "{SPARK_ROLE_ARN}"
      },
      "Action": "s3:ListBucket",
      "Resource": "arn:aws:s3:::tecton-{DEPLOYMENT_NAME}"
    }
  ]
}
```
2. Make sure the Object Ownership Setting for the "tecton-{DEPLOYMENT_NAME}" S3 Bucket is set to ACLs Disabled (Bucket Owner Enforced). This allows the AWS account with the S3 bucket to automatically own and have full control over objects written by the Databricks AWS Account.
  - In the AWS console open the S3 bucket configuration for the tecton-${DEPLOYMENT_NAME} bucket
    - https://s3.console.aws.amazon.com/s3/buckets/tecton-${DEPLOYMENT_NAME}?region=${AWS_REGION}&tab=permissions
  - Navigate to the permissions tab (if not already there)
  - Under the Edit Object Ownership section, click the Edit button
  - Select ACLs disabled (recommended)
  - This should change Object Ownership to Bucket owner enforced
  - Click Save Changes

Notify the Tecton team

Once you've completed the above setup, notify Tecton that you are ready to connect to AWS resources in your account.

Please include the following information: the ARN and External ID of the role you created for the Tecton control plane.

Was this page helpful?

Happy React is loading...

Configuring Databricks

Connecting Tecton to Databricks​

Configuring AWS resources​

Create a Tecton S3 Bucket​

Create a cross-account role for the Tecton Control Plane​

Create an AWS Policy for Spark​

Cross-Account Databricks Configuration​

Notify the Tecton team​