Service Foundry
Blog Infra SSO Observability Backend Big Data
Young Gyu Kim <credemol@gmail.com>

Apache Airflow - How to use SparkKubernetesOperator

Overview

This document provides how to install and use the Spark Operator on Kubernetes.

Installation

Add the Spark Operator Helm repository

We are going to use 'spark-operator' provided by Kubeflow. First, check you have added the old version of spark-operator repository provided by Google. Then remove it.

remove old Helm repository
# check Helm repositories
$ helm repo list | grep spark-operator

spark-operator          https://googlecloudplatform.github.io/spark-on-k8s-operator

$ helm repo remove spark-operator

Now, add the new Helm repository provided by Kubeflow.

add new Helm repository
$ helm repo add spark-operator https://kubeflow.github.io/spark-operator
$ helm repo update

$ helm repo list | grep spark-operator

spark-operator          https://kubeflow.github.io/spark-operator
download the values.yaml file.
$ helm show values spark-operator/spark-operator > values.yaml

custom-values.yaml

You can customize the values.yaml file to fit your needs. Here is an example of custom-values.yaml.

custom-values.yaml
controller:
  nodeSelector:
    agentpool: depnodes

webhook:
  nodeSelector:
    agentpool: depnodes

spark:
  jobNamespaces:
  - dep
  - spark-jobs
  - airflow
  - default

For spark.jobNamespace, you can specify the namespace where the Spark applications will be created.

Install the Spark Operator

$ helm install spark-operator spark-operator/spark-operator --namespace spark-operator --create-namespace -f custom-values.yaml

NAME: spark-operator
LAST DEPLOYED: Thu Dec  5 11:36:33 2024
NAMESPACE: spark-operator
STATUS: deployed
REVISION: 1
TEST SUITE: None

Uninstall the Spark Operator

$ helm uninstall spark-operator -n spark-operator

Create an example application

Spark-Operator examples can be found in the examples directory of the Spark-Operator repository.

examples/spark-pi.yaml
#
# Copyright 2017 Google LLC
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     https://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

apiVersion: sparkoperator.k8s.io/v1beta2
kind: SparkApplication
metadata:
  name: spark-pi
#  namespace: default
  namespace: spark-jobs
spec:
  type: Scala
  mode: cluster
  image: spark:3.5.3
  imagePullPolicy: IfNotPresent
  mainClass: org.apache.spark.examples.SparkPi
  mainApplicationFile: local:///opt/spark/examples/jars/spark-examples.jar
  arguments:
    - "5000"
  sparkVersion: 3.5.3
  driver:
    labels:
      version: 3.5.3
    cores: 1
    memory: 512m
    serviceAccount: spark-operator-spark
  executor:
    labels:
      version: 3.5.3
    instances: 1
    cores: 1
    memory: 512m

I just changed the namespace to 'spark-jobs' in the example file.

$ kubectl get namespace spark-jobs || kubectl create namespace spark-jobs

# Create an example Spark application in the spark-jobs namespace
$ kubectl apply -f examples/spark-pi.yaml -n spark-jobs

Verify the Spark application

To verify the Spark application, you can check the logs of the driver pod.

$ kubectl get pods -n spark-jobs
$ kubectl logs -f spark-pi-driver -n spark-jobs

Upgrade the Spark Operator

$ helm upgrade spark-operator spark-operator/spark-operator --namespace spark-operator -f custom-values.yaml

Working with Airflow

Spark Kubernetes Operator