aws-k8s-tester
is a set of utilities and libraries for "testing" Kubernetes on AWS.
- Implements
test-infra/kubetest2
interface. - Uses AWS CloudFormation for resource creation.
- Supports automatic rollback and resource deletion.
- Flexible add-on support via environmental variables.
- Extensible as a Go package;
eks.Tester.Up
to create EKS. - Performance tests suites.
The main goal is to create "temporary" EC2 instances or EKS clusters for "testing" purposes:
- Upstream conformance tests
- CNI plugin conformance tests
- AppMesh scalability testing
https://github.com/aws/aws-k8s-tester/releases
Make sure AWS credential is located in your machine:
# confirm credential is valid
aws sts get-caller-identity --query Arn --output text
See the following for more fields:
- https://github.com/aws/aws-k8s-tester/blob/master/eksconfig/README.md
- https://pkg.go.dev/github.com/aws/aws-k8s-tester/eksconfig?tab=doc
- https://github.com/aws/aws-k8s-tester/blob/master/eksconfig/default.yaml
# easiest way, use the defaults
# creates role, VPC, EKS cluster
rm -rf /tmp/${USER}-test-eks-prod*
aws-k8s-tester eks create cluster --enable-prompt=true -p /tmp/${USER}-test-prod-eks.yaml
aws-k8s-tester eks delete cluster --enable-prompt=true -p /tmp/${USER}-test-prod-eks.yaml
# advanced options can be set via environmental variables
# e.g. node groups, managed node groups, add-ons
rm -rf /tmp/${USER}-test-eks*
AWS_K8S_TESTER_EKS_PARTITION=aws \
AWS_K8S_TESTER_EKS_REGION=us-west-2 \
AWS_K8S_TESTER_EKS_S3_BUCKET_CREATE=true \
AWS_K8S_TESTER_EKS_S3_BUCKET_CREATE_KEEP=true \
AWS_K8S_TESTER_EKS_COMMAND_AFTER_CREATE_CLUSTER="aws eks describe-cluster --name GetRef.Name" \
AWS_K8S_TESTER_EKS_COMMAND_AFTER_CREATE_ADD_ONS="aws eks describe-cluster --name GetRef.Name" \
AWS_K8S_TESTER_EKS_PARAMETERS_ENCRYPTION_CMK_CREATE=true \
AWS_K8S_TESTER_EKS_PARAMETERS_ROLE_CREATE=true \
AWS_K8S_TESTER_EKS_PARAMETERS_VERSION=1.16 \
AWS_K8S_TESTER_EKS_PARAMETERS_VPC_CREATE=true \
AWS_K8S_TESTER_EKS_CLIENTS=5 \
AWS_K8S_TESTER_EKS_CLIENT_QPS=30 \
AWS_K8S_TESTER_EKS_CLIENT_BURST=20 \
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ROLE_CREATE=true \
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ASGS='{"GetRef.Name-ng-al2-cpu":{"name":"GetRef.Name-ng-al2-cpu","remote-access-user-name":"ec2-user","ami-type":"AL2_x86_64","image-id":"","image-id-ssm-parameter":"/aws/service/eks/optimized-ami/1.16/amazon-linux-2/recommended/image_id","instance-types":["c5.xlarge"],"volume-size":40,"asg-min-size":2,"asg-max-size":2,"asg-desired-capacity":2,"kubelet-extra-args":""},"GetRef.Name-ng-al2-gpu":{"name":"GetRef.Name-ng-al2-gpu","remote-access-user-name":"ec2-user","ami-type":"AL2_x86_64_GPU","image-id":"","image-id-ssm-parameter":"/aws/service/eks/optimized-ami/1.16/amazon-linux-2-gpu/recommended/image_id","instance-types":["p3.8xlarge"],"volume-size":40,"asg-min-size":1,"asg-max-size":1,"asg-desired-capacity":1,"kubelet-extra-args":""},"GetRef.Name-ng-bottlerocket":{"name":"GetRef.Name-ng-bottlerocket","remote-access-user-name":"ec2-user","ami-type":"BOTTLEROCKET_x86_64","image-id":"","image-id-ssm-parameter":"/aws/service/bottlerocket/aws-k8s-1.15/x86_64/latest/image_id","ssm-document-cfn-stack-name":"GetRef.Name-install-bottlerocket","ssm-document-name":"GetRef.Name-InstallBottlerocket","ssm-document-create":true,"ssm-document-commands":"enable-admin-container","ssm-document-execution-timeout-seconds":3600,"instance-types":["c5.xlarge"],"volume-size":40,"asg-min-size":2,"asg-max-size":2,"asg-desired-capacity":2}}' \
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_ROLE_CREATE=true \
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_MNGS='{"GetRef.Name-mng-al2-cpu":{"name":"GetRef.Name-mng-al2-cpu","remote-access-user-name":"ec2-user","release-version":"","ami-type":"AL2_x86_64","instance-types":["c5.xlarge"],"volume-size":40,"asg-min-size":2,"asg-max-size":2,"asg-desired-capacity":2},"GetRef.Name-mng-al2-gpu":{"name":"GetRef.Name-mng-al2-gpu","remote-access-user-name":"ec2-user","release-version":"","ami-type":"AL2_x86_64_GPU","instance-types":["p3.8xlarge"],"volume-size":40,"asg-min-size":1,"asg-max-size":1,"asg-desired-capacity":1}}' \
AWS_K8S_TESTER_EKS_ADD_ON_CONFORMANCE_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_CSI_EBS_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_KUBERNETES_DASHBOARD_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_PROMETHEUS_GRAFANA_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_APP_MESH_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_NLB_HELLO_WORLD_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_ALB_2048_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_JOBS_PI_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_JOBS_ECHO_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_CRON_JOBS_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_CSRS_LOCAL_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_CONFIG_MAPS_LOCAL_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_SECRETS_LOCAL_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_WORDPRESS_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_JUPYTER_HUB_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_HOLLOW_NODES_LOCAL_ENABLE=true \
AWS_K8S_TESTER_EKS_ADD_ON_CLUSTER_LOADER_LOCAL_ENABLE=true \
aws-k8s-tester eks create cluster --enable-prompt=true -p /tmp/${USER}-test-eks.yaml
<<COMMENT
# to delete
aws-k8s-tester eks delete cluster --enable-prompt=true -p /tmp/${USER}-test-eks.yaml
# run "eks create config" to check/edit configuration file first
aws-k8s-tester eks create config -p /tmp/${USER}-test-eks.yaml
# run the following command with those envs overwrites configuration, and create
aws-k8s-tester eks create cluster --enable-prompt=true -p /tmp/${USER}-test-eks.yaml
COMMENT
<<COMMENT
ACCOUNT_ID=$(aws sts get-caller-identity --query Account --output text);
echo ${ACCOUNT_ID}
CLUSTER_ARN=arn:aws:eks:us-west-2:${ACCOUNT_ID}:cluster/${USER}-test-eks
echo ${CLUSTER_ARN}
# to assign a non-random cluster name
# if empty, name is auto-generated
AWS_K8S_TESTER_EKS_NAME=${USER}-test-eks \
# to create/delete a S3 bucket for test artifacts
AWS_K8S_TESTER_EKS_S3_BUCKET_CREATE=true \
# to reuse an existing S3 bucket
AWS_K8S_TESTER_EKS_S3_BUCKET_CREATE=false \
AWS_K8S_TESTER_EKS_S3_BUCKET_NAME=${BUCKET_NAME} \
# to automatically create EC2 key-pair
AWS_K8S_TESTER_EKS_REMOTE_ACCESS_KEY_CREATE=true \
# to reuse an existing EC2 key-pair
AWS_K8S_TESTER_EKS_REMOTE_ACCESS_KEY_CREATE=false \
AWS_K8S_TESTER_EKS_REMOTE_ACCESS_KEY_NAME=${KEY_NAME} \
AWS_K8S_TESTER_EKS_REMOTE_ACCESS_PRIVATE_KEY_PATH=${KEY_PATH} \
# to reuse an existing role for "EKS cluster"
AWS_K8S_TESTER_EKS_PARAMETERS_ROLE_CREATE=false \
AWS_K8S_TESTER_EKS_PARAMETERS_ROLE_ARN=${CLUSTER_ROLE_ARN} \
# to reuse an existing VPC
AWS_K8S_TESTER_EKS_PARAMETERS_VPC_CREATE=false \
AWS_K8S_TESTER_EKS_PARAMETERS_VPC_ID=${VPC_ID} \
# to reuse an existing role for "Node Group"
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ROLE_CREATE=false \
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ROLE_ARN=${NG_ROLE_ARN} \
# to reuse an existing role for "Managed Node Group"
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_ROLE_CREATE=false \
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_ROLE_ARN=${MNG_ROLE_ARN} \
# to reuse an existing role for "Fargate"
AWS_K8S_TESTER_EKS_ADD_ON_FARGATE_ROLE_CREATE=false \
AWS_K8S_TESTER_EKS_ADD_ON_FARGATE_ROLE_ARN=${FARGATE_ROLE_ARN} \
# to user ${USER} in node groups
AWS_K8S_TESTER_EKS_ADD_ON_NODE_GROUPS_ASGS={\"${USER}-test-eks-ng-al2-cpu\":{\"name\":\"${USER}-test-eks-ng-al2-cpu\",\"remote-access-user-name\":\"ec2-user\",\"ami-type\":\"AL2_x86_64\",\"image-id-ssm-parameter\":\"/aws/service/eks/optimized-ami/1.15/amazon-linux-2/recommended/image_id\",\"instance-types\":[\"c5.xlarge\"],\"volume-size\":40,\"asg-min-size\":1,\"asg-max-size\":1,\"asg-desired-capacity\":1,\"kubelet-extra-args\":\"\"},\"${USER}-test-eks-ng-bottlerocket\":{\"name\":\"${USER}-test-eks-ng-bottlerocket\",\"remote-access-user-name\":\"ec2-user\",\"ami-type\":\"BOTTLEROCKET_x86_64\",\"image-id-ssm-parameter\":\"/aws/service/bottlerocket/aws-k8s-1.15/x86_64/latest/image_id\",\"ssm-document-cfn-stack-name\":\"${USER}-install-bottle-rocket\",\"ssm-document-name\":\"${USER}InstallBottleRocket\",\"ssm-document-create\":true,\"ssm-document-commands\":\"enable-admin-container\",\"ssm-document-execution-timeout-seconds\":3600,\"instance-types\":[\"c5.xlarge\"],\"volume-size\":40,\"asg-min-size\":1,\"asg-max-size\":1,\"asg-desired-capacity\":1}} \
AWS_K8S_TESTER_EKS_ADD_ON_MANAGED_NODE_GROUPS_MNGS={\"${USER}-test-eks-mng-al2-cpu\":{\"name\":\"${USER}-test-eks-mng-al2-cpu\",\"remote-access-user-name\":\"ec2-user\",\"release-version\":\"\",\"ami-type\":\"AL2_x86_64\",\"instance-types\":[\"c5.xlarge\"],\"volume-size\":40,\"asg-min-size\":1,\"asg-max-size\":1,\"asg-desired-capacity\":1}} \
COMMENT
This will create an EKS cluster with a worker node (takes about 20 minutes).
Once cluster is created, check cluster state using AWS CLI:
aws eks describe-cluster \
--name ${EKS_CLUSTER_NAME} \
--query cluster.status
"ACTIVE"
Cluster states are persisted on disk and S3 bucket.
EKS tester uses this file to record status.
cat /tmp/config.yaml
# or
less +FG /tmp/config.yaml
Make sure AWS credential is located in your machine:
# confirm credential is valid
aws sts get-caller-identity --query Arn --output text
See the following for more fields:
- https://github.com/aws/aws-k8s-tester/blob/master/ec2config/README.md
- https://pkg.go.dev/github.com/aws/aws-k8s-tester/ec2config?tab=doc
- https://github.com/aws/aws-k8s-tester/blob/master/ec2config/default.yaml
# easiest way, use the defaults
# creates role, VPC, EC2 ASG
rm -rf /tmp/${USER}-test-ec2*
ec2-utils create instances --enable-prompt=true -p /tmp/${USER}-test-ec2.yaml
ec2-utils delete instances --enable-prompt=true -p /tmp/${USER}-test-ec2.yaml
# advanced options can be set via environmental variables
rm -rf /tmp/${USER}-test-ec2*
AWS_K8S_TESTER_EC2_ON_FAILURE_DELETE=true \
AWS_K8S_TESTER_EC2_PARTITION=aws \
AWS_K8S_TESTER_EC2_REGION=us-west-2 \
AWS_K8S_TESTER_EC2_S3_BUCKET_CREATE=true \
AWS_K8S_TESTER_EC2_S3_BUCKET_CREATE_KEEP=true \
AWS_K8S_TESTER_EC2_REMOTE_ACCESS_KEY_CREATE=true \
AWS_K8S_TESTER_EC2_ASGS_FETCH_LOGS=true \
AWS_K8S_TESTER_EC2_ASGS='{"GetRef.Name-al2-cpu":{"name":"GetRef.Name-al2-cpu","remote-access-user-name":"ec2-user","ami-type":"AL2_x86_64","image-id-ssm-parameter":"/aws/service/ami-amazon-linux-latest/amzn2-ami-hvm-x86_64-gp2","instance-types":["c5.xlarge"],"volume-size":40,"asg-min-size":1,"asg-max-size":1,"asg-desired-capacity":1},"GetRef.Name-bottlerocket":{"name":"GetRef.Name-bottlerocket","remote-access-user-name":"ec2-user","ami-type":"BOTTLEROCKET_x86_64","image-id-ssm-parameter":"/aws/service/bottlerocket/aws-k8s-1.15/x86_64/latest/image_id","ssm-document-cfn-stack-name":"GetRef.Name-install-bottlerocket","ssm-document-name":"GetRef.Name-install-bottlerocket","ssm-document-create":true,"ssm-document-commands":"enable-admin-container","ssm-document-execution-timeout-seconds":3600,"instance-types":["c5.xlarge"],"volume-size":40,"asg-min-size":1,"asg-max-size":1,"asg-desired-capacity":1}}' \
AWS_K8S_TESTER_EC2_ROLE_CREATE=true \
AWS_K8S_TESTER_EC2_VPC_CREATE=true \
ec2-utils create instances --enable-prompt=true -p /tmp/${USER}-test-ec2.yaml
<<COMMENT
# to delete
ec2-utils delete instances --enable-prompt=true -p /tmp/${USER}-test-ec2.yaml
# run "ec2 create config" to check/edit configuration file first
ec2-utils create config -p /tmp/${USER}-test-ec2.yaml
ec2-utils create instances -p /tmp/${USER}-test-ec2.yaml
# run the following command with those envs overwrites configuration, and create
ec2-utils create instances --enable-prompt=true -p /tmp/${USER}-test-ec2.yaml
COMMENT
<<COMMENT
# to config a fixed name for EC2 ASG
AWS_K8S_TESTER_EC2_NAME=${NAME} \
# to create/delete a S3 bucket for test artifacts
AWS_K8S_TESTER_EC2_S3_BUCKET_CREATE=true \
# to reuse an existing S3 bucket
AWS_K8S_TESTER_EC2_S3_BUCKET_CREATE=false \
AWS_K8S_TESTER_EC2_S3_BUCKET_NAME=${BUCKET_NAME} \
# to automatically create EC2 key-pair
AWS_K8S_TESTER_EC2_REMOTE_ACCESS_KEY_CREATE=true \
# to reuse an existing EC2 key-pair
AWS_K8S_TESTER_EC2_REMOTE_ACCESS_KEY_CREATE=false \
AWS_K8S_TESTER_EC2_REMOTE_ACCESS_KEY_NAME=${KEY_NAME} \
AWS_K8S_TESTER_EC2_REMOTE_ACCESS_PRIVATE_KEY_PATH=${KEY_PATH} \
# to reuse an existing role
AWS_K8S_TESTER_EC2_ROLE_CREATE=false \
AWS_K8S_TESTER_EC2_ROLE_ARN=${ROLE_ARN} \
# to reuse an existing VPC
AWS_K8S_TESTER_EC2_VPC_CREATE=false \
AWS_K8S_TESTER_EC2_VPC_ID=${VPC_ID} \
# to use ${USER}
AWS_K8S_TESTER_EC2_ASGS={\"${USER}-test-ec2-al2-cpu\":{\"name\":\"${USER}-test-ec2-al2-cpu\",\"remote-access-user-name\":\"ec2-user\",\"ami-type\":\"AL2_x86_64\",\"image-id-ssm-parameter\":\"/aws/service/ami-amazon-linux-latest/amzn2-ami-hvm-x86_64-gp2\",,\"instance-types\":[\"c5.xlarge\"],\"volume-size\":40,\"asg-min-size\":1,\"asg-max-size\":1,\"asg-desired-capacity\":1},\"${USER}-test-ec2-bottlerocket\":{\"name\":\"${USER}-test-ec2-bottlerocket\",\"remote-access-user-name\":\"ec2-user\",\"ami-type\":\"BOTTLEROCKET_x86_64\",\"image-id-ssm-parameter\":\"/aws/service/bottlerocket/aws-k8s-1.15/x86_64/latest/image_id\",\"ssm-document-cfn-stack-name\":\"${USER}-install-bottlerocket\",\"ssm-document-name\":\"${USER}InstallBottleRocket\",\"ssm-document-create\":true,\"ssm-document-commands\":\"enable-admin-container\",\"ssm-document-execution-timeout-seconds\":3600,,\"instance-types\":[\"c5.xlarge\"],\"volume-size\":40,\"asg-min-size\":1,\"asg-max-size\":1,\"asg-desired-capacity\":1}} \
COMMENT
Install eks-utils
from https://github.com/aws/aws-k8s-tester/releases.
AWS_K8S_TESTER_VERSION=${LATEST_RELEASE_VERSION}
DOWNLOAD_URL=https://github.com/aws/aws-k8s-tester/releases/download
rm -rf /tmp/aws-k8s-tester
rm -rf /tmp/eks-utils
if [[ "${OSTYPE}" == "linux"* ]]; then
curl -L ${DOWNLOAD_URL}/${AWS_K8S_TESTER_VERSION}/eks-utils-${AWS_K8S_TESTER_VERSION}-linux-amd64 -o /tmp/eks-utils
elif [[ "${OSTYPE}" == "darwin"* ]]; then
curl -L ${DOWNLOAD_URL}/${AWS_K8S_TESTER_VERSION}/eks-utils-${AWS_K8S_TESTER_VERSION}-darwin-amd64 -o /tmp/eks-utils
fi
chmod +x /tmp/eks-utils
/tmp/eks-utils version
eks-utils apis
helps with API deprecation (e.g. https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.16.md#deprecations-and-removals).
WARNING: kubectl
internally converts API versions in the response (see kubernetes/issues#58131
). Which means kubectl get
output may have different API versions than the one persisted in etcd
. Upstream Kubernetes recommends upgrading deprecated API with get and put:
the simplest approach is to get/put every object after upgrades. objects that don't need migration will no-op (they won't even increment resourceVersion in etcd). objects that do need migration will persist in the new preferred storage version
Which means there's no way in client-side to find all resources created with deprecated API groups. The only way to ensure API group upgrades is list all resources, and execute get and put with the latest API group version. If the resource has already latest API version, it will be no-op. Otherwise, it will upgrade to the latest API version.
eks-utils apis
will help with the list calls with proper pagination and generate get and put scripts for the cluster:
# to check supported API groups from current kube-apiserver
eks-utils apis \
--kubeconfig /tmp/kubeconfig.yaml \
supported
# to write API upgrade/rollback scripts and YAML files in "/tmp/eks-utils"
#
# make sure to set proper "--batch-limit" and "--batch-interval"
# to not overload EKS master; if it's set too high, it can affect
# production workloads slowing down kube-apiserver
rm -rf /tmp/eks-utils-resources
eks-utils apis \
--kubeconfig /tmp/kubeconfig.yaml \
--enable-prompt \
deprecate \
--batch-limit 10 \
--batch-interval 2s \
--dir /tmp/eks-utils-resources
# this command does not apply or create any resources
# it only lists the resources that need be upgraded
# if there's any resources that needs upgrade,
# it writes patched YAML file, original YAML file,
# bash scripts to update and rollback
find /tmp/eks-utils-resources
etcd-utils k8s list
helps with API deprecation (e.g. https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG/CHANGELOG-1.16.md#deprecations-and-removals).
WARNING: kubectl
internally converts API versions in the response (see kubernetes/issues#58131
). Which means kubectl get
output may have different API versions than the one persisted in etcd
. Upstream Kubernetes recommends upgrading deprecated API with get and put:
the simplest approach is to get/put every object after upgrades. objects that don't need migration will no-op (they won't even increment resourceVersion in etcd). objects that do need migration will persist in the new preferred storage version
To minimize the impact of list calls, etcd-utils k8s list
reads keys with leadership election and pagination; only a single worker can run at a time.
# to list all deployments with etcd pagination + k8s decoder
etcd-utils k8s \
--endpoints http://localhost:2379 \
list \
--prefixes /registry/deployments \
--output /tmp/etcd-utils-k8s-list.output.yaml
# or ".json"