Call Amazon EKS with Step Functions - Amazon Step Functions
Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Call Amazon EKS with Step Functions

Step Functions can control certain Amazon services directly from Amazon States Language (ASL). To learn more, see Working with other services and Pass parameters to a service API.

How the Optimized Amazon EKS integration is different than the Amazon EKS Amazon SDK integration

For information about how to configure IAM permissions when using Step Functions with other Amazon services, see IAM Policies for integrated services.

Step Functions provides two types of service integration APIs for integrating with Amazon Elastic Kubernetes Service. One lets you use the Amazon EKS APIs to create and manage an Amazon EKS cluster. The other lets you interact with your cluster using the Kubernetes API and run jobs as part of your application’s workflow. You can use the Kubernetes API integrations with Amazon EKS clusters created using Step Functions, with Amazon EKS clusters created by the eksctl tool or the Amazon EKS console, or similar methods. For more information, see Creating an Amazon EKS cluster in the Amazon EKS User Guide.

Note

The Step Functions EKS integration supports only Kubernetes APIs with public endpoint access. By default, EKS clusters API server endpoints have public access. For more information, see Amazon EKS cluster endpoint access control in the Amazon EKS User Guide.

Step Functions does not terminate an Amazon EKS cluster automatically if execution is stopped. If your state machine stops before your Amazon EKS cluster has terminated, your cluster may continue running indefinitely, and can accrue additional charges. To avoid this, ensure that any Amazon EKS cluster you create is terminated properly. For more information, see:

Note

There is a quota for the maximum input or result data size for a task in Step Functions. This restricts you to 256 KB of data as a UTF-8 encoded string when you send to, or receive data from, another service. See Quotas related to state machine executions.

Kubernetes API integrations

Step Functions supports the following Kubernetes APIs:

RunJob

The eks:runJob service integration allows you to run a job on your Amazon EKS cluster. The eks:runJob.sync variant allows you to wait for the job to complete, and, optionally retrieve logs.

Your Kubernetes API server must grant permissions to the IAM role used by your state machine. For more information, see Permissions.

For the Run a Job (.sync) pattern, the status of the job is determined by polling. Step Functions initially polls at a rate of approximately 1 poll per minute. This rate eventually slows to approximately 1 poll every 5 minutes. If you require more frequent polling, or require more control over the polling strategy, you can use the eks:call integration to query the status of the job.

The eks:runJob integration is specific to batch/v1 Kubernetes Jobs. For more information, see Jobs in the Kubernetes documentation. If you want to manage other Kubernetes resources, including custom resources, use the eks:call service integration. You can use Step Functions to build polling loops, as demonstrated in the Poll for Job Status (Lambda, Amazon Batch) sample project.

Supported parameters include:

  • ClusterName: The name of the Amazon EKS cluster you want to call.

    • Type: String

    • Required: yes

  • CertificateAuthority: The Base64-encoded certificate data required to communicate with your cluster. You can obtain this value from the Amazon EKS console or by using the Amazon EKS DescribeCluster API.

    • Type: String

    • Required: yes

  • Endpoint: The endpoint URL for your Kubernetes API server. You can obtain this value from the Amazon EKS console or by using the Amazon EKS DescribeCluster API.

    • Type: String

    • Required: yes

  • Namespace: The namespace in which to run the job. If not provided, the namespace default is used.

    • Type: String

    • Required: no

  • Job: The definition of the Kubernetes Job. See Jobs in the Kubernetes documentation.

    • Type: JSON or String

    • Required: yes

  • LogOptions: A set of options to control the optional retrieval of logs. Only applicable if the Run a Job (.sync) service integration pattern is used to wait for the completion of the job.

    • Type: JSON

    • Required: no

    • Logs are included in the response under the key logs. There may be multiple pods within the job, each with multiple containers.

      { ... "logs": { "pods": { "pod1": { "containers": { "container1": { "log": <log> }, ... } }, ... } }
    • Log retrieval is performed on a best-effort basis. If there is an error retrieving a log, in place of the log field there will be the fields error and cause.

  • LogOptions.RetrieveLogs: Enable log retrieval after the job completes. By default, logs are not retrieved.

    • Type: Boolean

    • Required: no

  • LogOptions.RawLogs: If RawLogs is set to true, logs will be returned as raw strings without attempting to parse them into JSON. By default, logs are deserialized into JSON if possible. In some cases such parsing can introduce unwanted changes, such as limiting the precision of numbers containing many digits.

    • Type: Boolean

    • Required: no

  • LogOptions.LogParameters: The Kubernetes API’s Read Log API supports query parameters to control log retrieval. For example, you can use tailLines or limitBytes to limit the size of retrieved logs and remain within the Step Functions data size quota. For more information, see the Read Log section of the Kubernetes API Reference.

    • Type: Map of String to List of Strings

    • Required: no

    • Example:

      "LogParameters": { "tailLines": [ "6" ] }

The following example includes a Task state that runs a job, waits for it to complete, then retrieves the job’s logs:

{ "StartAt": "Run a job on EKS", "States": { "Run a job on EKS": { "Type": "Task", "Resource": "arn:aws:states:::eks:runJob.sync", "Parameters": { "ClusterName": "MyCluster", "CertificateAuthority": "ANPAJ2UCCR6DPCEXAMPLE", "Endpoint": "https://AKIAIOSFODNN7EXAMPLE.yl4.us-east-1.eks.amazonaws.com", "LogOptions": { "RetrieveLogs": true }, "Job": { "apiVersion": "batch/v1", "kind": "Job", "metadata": { "name": "example-job" }, "spec": { "backoffLimit": 0, "template": { "metadata": { "name": "example-job" }, "spec": { "containers": [ { "name": "pi-2000", "image": "perl", "command": [ "perl" ], "args": [ "-Mbignum=bpi", "-wle", "print bpi(2000)" ] } ], "restartPolicy": "Never" } } } } }, "End": true } } }

Call

The eks:call service integration allows you to use the Kubernetes API to read and write Kubernetes resource objects via a Kubernetes API endpoint.

Your Kubernetes API server must grant permissions to the IAM role used by your state machine. For more information, see Permissions.

For more information about the available operations, see the Kubernetes API Reference.

Supported parameters for Call include:

  • ClusterName: The name of the Amazon EKS cluster you want to call.

    • Type: String

    • Required: Yes

  • CertificateAuthority: The Base64-encoded certificate data required to communicate with your cluster. You can obtain this value from the Amazon EKS console or by using the Amazon EKS DescribeCluster API.

    • Type: String

    • Required: Yes

  • Endpoint: The endpoint URL for your Kubernetes API server. You can find this value on the Amazon EKS console or by using Amazon EKS’ DescribeCluster API.

    • Type: String

    • Required: Yes

  • Method: The HTTP method of your request. One of: GET, POST, PUT, DELETE, HEAD, or PATCH.

    • Type: String

    • Required: Yes

  • Path: The HTTP path of the Kubernetes REST API operation.

    • Type: String

    • Required: Yes

  • QueryParameters: The HTTP query parameters of the Kubernetes REST API operation.

    • Type: Map of String to List of Strings

    • Required: No

    • Example:

      "QueryParameters": { "labelSelector": [ "job-name=example-job" ] }
  • RequestBody: The HTTP message body of the Kubernetes REST API operation.

    • Type: JSON or String

    • Required: No

The following includes a Task state that uses eks:call to list the pods belonging to the job example-job.

{ "StartAt": "Call EKS", "States": { "Call EKS": { "Type": "Task", "Resource": "arn:aws:states:::eks:call", "Parameters": { "ClusterName": "MyCluster", "CertificateAuthority": "ANPAJ2UCCR6DPCEXAMPLE", "Endpoint": "https://444455556666.yl4.us-east-1.eks.amazonaws.com", "Method": "GET", "Path": "/api/v1/namespaces/default/pods", "QueryParameters": { "labelSelector": [ "job-name=example-job" ] } }, "End": true } } }

The following includes a Task state that uses eks:call to delete the job example-job, and sets the propagationPolicy to ensure the job's pods are also deleted.

{ "StartAt": "Call EKS", "States": { "Call EKS": { "Type": "Task", "Resource": "arn:aws:states:::eks:call", "Parameters": { "ClusterName": "MyCluster", "CertificateAuthority": "ANPAJ2UCCR6DPCEXAMPLE", "Endpoint": "https://444455556666.yl4.us-east-1.eks.amazonaws.com", "Method": "DELETE", "Path": "/apis/batch/v1/namespaces/default/jobs/example-job", "QueryParameters": { "propagationPolicy": [ "Foreground" ] } }, "End": true } } }

Supported Amazon EKS APIs

Supported Amazon EKS APIs and syntax include:

  • CreateCluster

    • Request syntax

    • Response syntax

      When an Amazon EKS cluster is created using the eks:createCluster service integration, the IAM role is added to the Kubernetes RBAC authorization table as the administrator (with system:masters permissions). Initially, only that IAM entity can make calls to the Kubernetes API server. For more information, see:

      Amazon EKS uses service-linked roles which contain the permissions Amazon EKS requires to call other services on your behalf. If these service-linked roles do not exist in your account already, you must add the iam:CreateServiceLinkedRole permission to the IAM role used by Step Functions. For more information, see Using Service-Linked Roles in the Amazon EKS User Guide.

      The IAM role used by Step Functions must have iam:PassRole permissions to pass the cluster IAM role to Amazon EKS. For more information, see Amazon EKS cluster IAM role in the Amazon EKS User Guide.

  • DeleteCluster

  • CreateFargateProfile

    • Request syntax

    • Response syntax

      Amazon EKS uses service-linked roles which contain the permissions Amazon EKS requires to call other services on your behalf. If these service-linked roles do not exist in your account already, you must add the iam:CreateServiceLinkedRole permission to the IAM role used by Step Functions. For more information, see Using Service-Linked Roles in the Amazon EKS User Guide.

      Amazon EKS on Fargate may not be available in all regions. For information on region availability, see the section on Fargate in the Amazon EKS User Guide.

      The IAM role used by Step Functions must have iam:PassRole permissions to pass the pod execution IAM role to Amazon EKS. For more information, see Pod execution role in the Amazon EKS User Guide.

  • DeleteFargateProfile

  • CreateNodegroup

    • Request syntax

    • Response syntax

      Amazon EKS uses service-linked role whichs contain the permissions Amazon EKS requires to call other services on your behalf. If these service-linked roles do not exist in your account already, you must add the iam:CreateServiceLinkedRole permission to the IAM role used by Step Functions. For more information, see Using Service-Linked Roles in the Amazon EKS User Guide.

      The IAM role used by Step Functions must have iam:PassRole permissions to pass the node IAM role to Amazon EKS. For more information, see Using Service-Linked Roles in the Amazon EKS User Guide.

  • DeleteNodegroup

The following includes a Task that creates an Amazon EKS cluster.

{ "StartAt": "CreateCluster.sync", "States": { "CreateCluster.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:createCluster.sync", "Parameters": { "Name": "MyCluster", "ResourcesVpcConfig": { "SubnetIds": [ "subnet-053e7c47012341234", "subnet-027cfea4b12341234" ] }, "RoleArn": "arn:aws:iam::123456789012:role/MyEKSClusterRole" }, "End": true } } }

The following includes a Task state that deletes an Amazon EKS cluster.

{ "StartAt": "DeleteCluster.sync", "States": { "DeleteCluster.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:deleteCluster.sync", "Parameters": { "Name": "MyCluster" }, "End": true } } }

The following includes a Task state that creates a Fargate profile.

{ "StartAt": "CreateFargateProfile.sync", "States": { "CreateFargateProfile.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:createFargateProfile.sync", "Parameters": { "ClusterName": "MyCluster", "FargateProfileName": "MyFargateProfile", "PodExecutionRoleArn": "arn:aws:iam::123456789012:role/MyFargatePodExecutionRole", "Selectors": [{ "Namespace": "my-namespace", "Labels": { "my-label": "my-value" } }] }, "End": true } } }

The following includes a Task state that deletes a Fargate profile.

{ "StartAt": "DeleteFargateProfile.sync", "States": { "DeleteFargateProfile.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:deleteFargateProfile.sync", "Parameters": { "ClusterName": "MyCluster", "FargateProfileName": "MyFargateProfile" }, "End": true } } }

The following includes a Task state that creates a node group.

{ "StartAt": "CreateNodegroup.sync", "States": { "CreateNodegroup.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:createNodegroup.sync", "Parameters": { "ClusterName": "MyCluster", "NodegroupName": "MyNodegroup", "NodeRole": "arn:aws:iam::123456789012:role/MyNodeInstanceRole", "Subnets": ["subnet-09fb51df01234", "subnet-027cfea4b1234"] }, "End": true } } }

The following includes a Task state that deletes a node group.

{ "StartAt": "DeleteNodegroup.sync", "States": { "DeleteNodegroup.sync": { "Type": "Task", "Resource": "arn:aws:states:::eks:deleteNodegroup.sync", "Parameters": { "ClusterName": "MyCluster", "NodegroupName": "MyNodegroup" }, "End": true } } }

Permissions

When an Amazon EKS cluster is created using the eks:createCluster service integration, the IAM role is added to the Kubernetes RBAC authorization table as the administrator, with system:masters permissions. Initially, only that IAM entity can make calls to the Kubernetes API server. For example, you will not be able to use kubectl to interact with your Kubernetes API server, unless you assume the same role as your Step Functions state machine, or if you configure Kubernetes to grant permissions to additional IAM entities. For more information, see Managing users or IAM roles for your cluster in the Amazon EKS User Guide.

You can add permission for additional IAM entities, such as users or roles, by adding them to the aws-auth ConfigMap in the kube-system namespace. If you are creating your cluster from Step Functions, use the eks:call service integration.

The following includes a Task state that creates an aws-auth ConfigMap and grants system:masters permission to the user arn:aws-cn:iam::123456789012:user/my-user and the IAM role arn:aws-cn:iam::123456789012:role/my-role.

{ "StartAt": "Add authorized user", "States": { "Add authorized user": { "Type": "Task", "Resource": "arn:aws-cn:states:::eks:call", "Parameters": { "ClusterName": "MyCluster", "CertificateAuthority": "LS0tLS1CRUd...UtLS0tLQo=", "Endpoint": "https://444455556666.yl4.us-east-1.eks.amazonaws.com", "Method": "POST", "Path": "/api/v1/namespaces/kube-system/configmaps", "RequestBody": { "apiVersion": "v1", "kind": "ConfigMap", "metadata": { "name": "aws-auth", "namespace": "kube-system" }, "data": { "mapUsers": "[{ \"userarn\": \"arn:aws:iam::123456789012:user/my-user\", \"username\": \"my-user\", \"groups\": [ \"system:masters\" ] } ]", "mapRoles": "[{ \"rolearn\": \"arn:aws:iam::123456789012:role/my-role\", \"username\": \"my-role\", \"groups\": [ \"system:masters\" ] } ]" } } }, "End": true } }
Note

You may see the ARN for an IAM role displayed in a format that includes the path /service-role/, such as arn:aws:iam::123456789012:role/service-role/my-role. This service-role path token should not be included when listing the role in aws-auth.

When your cluster is first created the aws-auth ConfigMap will not exist, but will be added automatically if you create a Fargate profile. You can retrieve the current value of aws-auth, add the additional permissions, and PUT a new version. It is usually easier to create aws-auth before the Fargate profile.

If your cluster was created outside of Step Functions, you can configure kubectl to communicate with your Kubernetes API server. Then, create a new aws-auth ConfigMap using kubectl apply -f aws-auth.yaml or edit one that already exists using kubectl edit -n kube-system configmap/aws-auth. For more information, see:

If your IAM role does not have sufficient permissions in Kubernetes, the eks:call or eks:runJob service integrations will fail with the following error:

Error: EKS.401 Cause: { "ResponseBody": { "kind": "Status", "apiVersion": "v1", "metadata": {}, "status": "Failure", "message": "Unauthorized", "reason": "Unauthorized", "code": 401 }, "StatusCode": 401, "StatusText": "Unauthorized" }