Spark Operator running on Amazon EKS IPv6
This example showcases the usage of Spark Operator running on Amazon EKS in IPv6 mode. the idea is to show and demonstarte running spark workloads on EKS IPv6 cluster.
Deploy the EKS Cluster with all the add-ons and infrastructure needed to test this example
The Terraform blueprint will provision the following resources required to run Spark Jobs with open source Spark Operator on Amazon EKS IPv6
- A Dual Stack Amazon Virtual Private Cloud (Amazon VPC) with 3 Private Subnets and 3 Public Subnets
- An Internet gateway for Public Subnets, NAT Gateway for Private Subnets and Egress-only Internet gateway
- An Amazon EKS cluster in IPv6 mode (version 1.30)
- Amazon EKS core-managed node group used to host some of the add-ons that we’ll provision on the cluster
- Deploys Spark-k8s-operator, Apache Yunikorn, Karpenter, Prometheus and Grafana server.
Prerequisites
Ensure that you have installed the following tools on your machine.
Before installing the cluster create a EKS IPv6 CNI policy. Follow the instructions from the link: AmazonEKS_CNI_IPv6_Policy
Clone the repository
git clone https://github.com/awslabs/data-on-eks.git
cd data-on-eks
export DOEKS_HOME=$(pwd)
Initialize Terraform
Navigate into the example directory and run the initialization script install.sh.
cd ${DOEKS_HOME}/analytics//terraform/spark-eks-ipv6/
chmod +x install.sh
./install.sh
Export Terraform Outputs
export CLUSTER_NAME=$(terraform output -raw cluster_name)
export AWS_REGION=$(terraform output -raw region)
export S3_BUCKET=$(terraform output -raw s3_bucket_id_spark_event_logs_example_data)
The S3_BUCKET variable that holds the name of the bucket created during the install. This bucket will be used in later examples to store output data.
Update kubeconfig
Update the kubeconfig to verify the deployment.
aws eks --region $AWS_REGION update-kubeconfig --name $CLUSTER_NAME
Verify the deployment
Examine the IP addresses assigned to the cluster nodes and the pods. You will notice that both have IPv6 addresses allocated.
kubectl get node -o custom-columns='NODE_NAME:.metadata.name,INTERNAL-IP:.status.addresses[?(@.type=="InternalIP")].address'
NODE_NAME INTERNAL-IP
ip-10-1-0-212.us-west-2.compute.internal 2600:1f13:520:1303:c87:4a71:b9ea:417c
ip-10-1-26-137.us-west-2.compute.internal 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa
ip-10-1-46-28.us-west-2.compute.internal 2600:1f13:520:1305:5ee5:b994:c0c2:e4da
kubectl get pods -A -o custom-columns='NAME:.metadata.name,NodeIP:.status.hostIP,PodIP:status.podIP'
NAME NodeIP PodIP
....
karpenter-5fd95dffb8-l8j26 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::
karpenter-5fd95dffb8-qpv55 2600:1f13:520:1303:c87:4a71:b9ea:417c 2600:1f13:520:1303:60ac::
kube-prometheus-stack-grafana-9f5c9d8fc-zgn98 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::a
kube-prometheus-stack-kube-state-metrics-98c74d866-56275 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::9
kube-prometheus-stack-operator-67df8bc57d-2d8jh 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::b
kube-prometheus-stack-prometheus-node-exporter-5qrqs 2600:1f13:520:1303:c87:4a71:b9ea:417c 2600:1f13:520:1303:c87:4a71:b9ea:417c
kube-prometheus-stack-prometheus-node-exporter-hcpvk 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa
kube-prometheus-stack-prometheus-node-exporter-ztkdm 2600:1f13:520:1305:5ee5:b994:c0c2:e4da 2600:1f13:520:1305:5ee5:b994:c0c2:e4da
prometheus-kube-prometheus-stack-prometheus-0 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::7
spark-history-server-6c9f9d7cc4-xzj4c 2600:1f13:520:1305:5ee5:b994:c0c2:e4da 2600:1f13:520:1305:64b::1
spark-operator-84c6b48ffc-z2glj 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::5
spark-operator-webhook-init-kbl4s 2600:1f13:520:1305:5ee5:b994:c0c2:e4da 2600:1f13:520:1305:64b::2
yunikorn-admission-controller-d675f89c5-f2p47 2600:1f13:520:1303:c87:4a71:b9ea:417c 2600:1f13:520:1303:c87:4a71:b9ea:417c
yunikorn-scheduler-59d6879975-2rh4d 2600:1f13:520:1304:15b2:b8a3:7f63:cbfa 2600:1f13:520:1304:a79b::4
....
Execute Sample Spark job with Karpenter
Navigate to example directory and submit the Spark job.
cd ${DOEKS_HOME}/analytics/terraform/spark-eks-ipv6/examples/karpenter
kubectl apply -f pyspark-pi-job.yaml
Monitor the job status using the below command. You should see the new nodes triggered by the Karpenter.
kubectl get pods -n spark-team-a -w
Apache YuniKorn Gang Scheduling with NVMe based SSD disk for shuffle storage
Gang Scheduling Spark jobs using Apache YuniKorn and Spark Operator
cd ${DOEKS_HOME}/analytics/terraform/spark-eks-ipv6/examples/karpenter/nvme-yunikorn-gang-scheduling
Run the taxi-trip-execute.sh script with the following input. You will use the S3_BUCKET variable created earlier. Additionally, you must change YOUR_REGION_HERE with the region of your choice, us-west-2 for example.
This script will download some example taxi trip data and create duplicates of it in order to increase the size a bit. This will take a bit of time and will require a relatively fast internet connection.
${DOEKS_HOME}/analytics/scripts/taxi-trip-execute.sh ${S3_BUCKET} YOUR_REGION_HERE
Once our sample data is uploaded you can run the Spark job. You will need to replace the <S3_BUCKET> placeholders in this file with the name of the bucket created earlier. You can get that value by running echo $S3_BUCKET.
To do this automatically you can run the following, which will create a .old backup file and do the replacement for you.
sed -i.old s/\<S3_BUCKET\>/${S3_BUCKET}/g ./nvme-storage-yunikorn-gang-scheduling.yaml
Now that the bucket name is in place you can create the Spark job.
kubectl apply -f nvme-storage-yunikorn-gang-scheduling.yaml
Cleanup
This script will cleanup the environment using -target option to ensure all the resources are deleted in correct order.
cd ${DOEKS_HOME}/analytics/terraform/spark-eks-ipv6 && chmod +x cleanup.sh
./cleanup.sh
To avoid unwanted charges to your AWS account, delete all the AWS resources created during this deployment