Azure Databricks operator (for Kubernetes)

This project is experimental. Expect the API to change. It is not recommended for production environments.

Introduction

Kubernetes offers the facility of extending it's API through the concept of 'Operators' (Introducing Operators: Putting Operational Knowledge into Software). This repository contains the resources and code to deploy an Azure Databricks Operator for Kubernetes.

The Azure Databricks operator comprises two projects:

The golang application is a Kubernetes controller that watches Customer Resource Definitions (CRDs) that define a Databricks job, and,
A Python Flask App which sends commands to the Databricks clusters.

The Databricks operator is useful in situations where Kubernetes hosted applications wish to launch and use Databricks data engineering and machine learning tasks.

The project was built using

Prerequisites And Assumptions

You have the kubectl command line (kubectl CLI) installed.
You have acess to a Kubernetes cluster. It can be a local hosted Cluster like Minikube, Kind or, Docker for desktop installed localy with RBAC enabled. if you opt for Azure Kubernetes Service (AKS), you can use: az aks get-credentials --resource-group $RG_NAME --name $Cluster_NAME

To configure a local Kubernetes cluster on your machine

You need to make sure a kubeconfig file is configured.

kustomize is also needed

Basic commands to check your cluster

    kubectl config get-contexts
    kubectl cluster-info
    kubectl version
    kubectl get pods -n kube-system

How to use the operator

Documentation is a work in progress

Quick start

Download latest release.zip

wget https://github.com/microsoft/azure-databricks-operator/releases/latest/download/release.zip
unzip release.zip

Create the databricks-operator-system namespace

kubectl create namespace databricks-operator-system

Generate a databricks token, and create Kubernetes secrets with values for DATABRICKS_HOST and DATABRICKS_TOKEN

    kubectl  --namespace databricks-operator-system create secret generic dbrickssettings --from-literal=DatabricksHost="https://xxxx.azuredatabricks.net" --from-literal=DatabricksToken="xxxxx"

Apply the manifests for the CRD and Operator in release/config:

kubectl apply -f release/config

Create a test secret, you can pass the value of Kubernetes secrets into your notebook as Databricks secrets

kubectl create secret generic test --from-literal=my_secret_key="my_secret_value"

In Databricks, create a new Python Notebook called testnotebook in the root of your Workspace. Put the following in the first cell of the notebook:

run_name = dbutils.widgets.get("run_name")
secret_scope = run_name + "_scope"

secret_value = dbutils.secrets.get(scope=secret_scope, key="dbricks_secret_key") # this will come from a kubernetes secret
print(secret_value) # will be redacted

value = dbutils.widgets.get("flag")
print(value) # 'true'

Define your Notebook job and apply it

apiVersion: microsoft.k8s.io/v1beta1
kind: NotebookJob
metadata:
  annotations:
    microsoft.k8s.io/author: azkhojan@microsoft.com
  name: samplejob1
spec:
  notebookTask:
    notebookPath: "/testnotebook"
  timeoutSeconds: 500
  notebookSpec:
    "flag":  "true"
  notebookSpecSecrets:
  - secretName: "test"
    mapping: 
    - "secretKey": "my_secret_key"
      "outputKey": "dbricks_secret_key"
  notebookAdditionalLibraries: 
    - type: "maven"
      properties:
        coordinates: "com.microsoft.azure:azure-eventhubs-spark_2.11:2.3.9" # installs the azure event hubs library
  clusterSpec:
    sparkVersion: "5.2.x-scala2.11"
    nodeTypeId: "Standard_DS12_v2"
    numWorkers: 1

Check the NotebookJob and Operator pod

# list all notebook jobs
kubectl get notebookjob
# describe a notebook job
kubectl describe notebookjob samplejob1
# describe the operator pod
kubectl -n databricks-operator-system describe pod databricks-operator-controller-manager-0
# get logs from the manager container
kubectl -n databricks-operator-system logs databricks-operator-controller-manager-0 -c dbricks

Check the job ran with expected output in the Databricks UI.

Run Souce Code

Clone the repo
Install the NotebookJob CRD in the configured Kubernetes cluster folder ~/.kube/config, run kubectl apply -f databricks-operator/config/crds or make install -C databricks-operator

Create secrets for DATABRICKS_HOST and DATABRICKS_TOKEN

kubectl  --namespace databricks-operator-system create secret generic dbrickssettings --from-literal=DatabricksHost="https://xxxx.azuredatabricks.net" --from-literal=DatabricksToken="xxxxx"

Make sure your secret name is set correctly in databricks-operator/config/default/azure_databricks_api_image_patch.yaml

Deploy the controller in the configured Kubernetes cluster folder ~/.kube/config, run kustomize build databricks-operator/config | kubectl apply -f -
Change the NotebookJob name from sample1run1 to your desired name, set the Databricks notebook path and update the values in microsoft_v1beta2_notebookjob.yaml to reflect your Databricks environment
```
kubectl apply -f databricks-operator/config/samples/microsoft_v1beta2_notebookjob.yaml
```

How to extend the operator and build your own images

Updating the Databricks operator:

This Repo is generated by Kubebuilder.

To Extend the operator databricks-operator:

Run dep ensure to download dependencies. It doesn't show any progress bar and takes a while to download all of dependencies.
Update pkg\apis\microsoft\v1beta1\notebookjob_types.go.
Regenerate CRD make manifests.
Install updated CRD make install
Generate code make generate
Update operator pkg\controller\notebookjob\notebookjob_controller.go
Update tests and run make test
Build make build

Deploy

make docker-build IMG=azadehkhojandi/databricks-operator
make docker-push IMG=azadehkhojandi/databricks-operator
make deploy

Main Contributors

Jordan Knight Github, Linkedin
Paul Bouwer Github, Linkedin
Lace Lofranco Github, Linkedin
Allan Targino Github, Linkedin
Jason Goodselli Github, Linkedin
Craig Rodger Github, Linkedin
Justin Chizer Github, Linkedin
Azadeh Khojandi Github, Linkedin

Resources

Kubernetes on WSL

On windows command line run kubectl config view to find the values of [windows-user-name],[minikubeip],[port]

mkdir ~/.kube \
&& cp /mnt/c/Users/[windows-user-name]/.kube/config ~/.kube

kubectl config set-cluster minikube --server=https://<minikubeip>:<port> --certificate-authority=/mnt/c/Users/<windows-user-name>/.minikube/ca.crt
kubectl config set-credentials minikube --client-certificate=/mnt/c/Users/<windows-user-name>/.minikube/client.crt --client-key=/mnt/c/Users/<windows-user-name>/.minikube/client.key
kubectl config set-context minikube --cluster=minikube --user=minikub

More info:

Build pipelines

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

phanvinh0526 / azure-databricks-operator