Skip to content



Amazon EKS Pod Identity Webhook

This webhook is for mutating pods that will require AWS IAM access.


After version v0.3.0, --in-cluster=true no longer works and is deprecated. Please use --in-cluster=false and manage the cluster certificate with cert-manager or some other external certificate provisioning system. This is because certificates using the legacy-unknown signer are no longer signed when using the v1 certificates API.

EKS Walkthrough

  1. Create an OIDC provider in IAM for your cluster. You can find the OIDC discovery endpoint by describing your EKS cluster.
    aws eks describe-cluster --name $CLUSTER_NAME --query cluster.identity.oidc
    And enter "" as the client-id
  2. Create an IAM role for your pods and modify the trust policy to allow your pod's service account to use the role:
     "Version": "2012-10-17",
     "Statement": [
       "Effect": "Allow",
       "Principal": {
        "Federated": "arn:aws:iam::111122223333:oidc-provider/"
       "Action": "sts:AssumeRoleWithWebIdentity",
       "Condition": {
        "__doc_comment": "scope the role to the service account (optional)",
        "StringEquals": {
         "": "system:serviceaccount:default:my-serviceaccount"
        "__doc_comment": "scope the role to a namespace (optional)",
        "StringLike": {
         "": "system:serviceaccount:default:*"
  3. Modify your pod's service account to be annotated with the ARN of the role you want the pod to use
    apiVersion: v1
    kind: ServiceAccount
      name: my-serviceaccount
      namespace: default
      annotations: "arn:aws:iam::111122223333:role/s3-reader"
        # optional: Defaults to "" if not set ""
        # optional: When set to "true", adds AWS_STS_REGIONAL_ENDPOINTS env var
        #   to containers "true"
        # optional: Defaults to 86400 for expirationSeconds if not set
        #   Note: This value can be overwritten if specified in the pod 
        #         annotation as shown in the next step. "86400"
  4. All new pod pods launched using this Service Account will be modified to use IAM for pods. Below is an example pod spec with the environment variables and volume fields added by the webhook.
    apiVersion: v1
    kind: Pod
      name: my-pod
      namespace: default
        # optional: A comma-separated list of initContainers and container names
        #   to skip adding volumes and environment variables "init-first,sidecar"
        # optional: Defaults to 86400, or value specified in ServiceAccount
        #   annotation as shown in previous step, for expirationSeconds if not set "86400"
      serviceAccountName: my-serviceaccount
      - name: init-first
        image: container-image:version
      - name: sidecar
        image: container-image:version
      - name: container-name
        image: container-image:version
    ### Everything below is added by the webhook ###
        - name: AWS_DEFAULT_REGION
          value: us-west-2
        - name: AWS_REGION
          value: us-west-2
        - name: AWS_ROLE_ARN
          value: "arn:aws:iam::111122223333:role/s3-reader"
          value: "/var/run/secrets/"
          value: "regional"
        - mountPath: "/var/run/secrets/"
          name: aws-token
      - name: aws-token
          - serviceAccountToken:
              audience: ""
              expirationSeconds: 86400
              path: token

Usage with Windows container workloads

To ensure workloads are scheduled on windows nodes have the right environment variables, they must have a nodeSelector targeting windows it must run on. Workloads targeting windows nodes using nodeAffinity are currently not supported.

  nodeSelector: windows

Or for Kubernetes 1.14+

  nodeSelector: windows

Usage with non-root container user

When running a container with a non-root user, you need to give the container access to the token file by setting the fsGroup field in the securityContext object.


Usage of amazon-eks-pod-identity-webhook:
      --add_dir_header                   If true, adds the file directory to the header
      --alsologtostderr                  log to standard error as well as files
      --annotation-prefix string         The Service Account annotation to look for (default "")
      --aws-default-region string        If set, AWS_DEFAULT_REGION and AWS_REGION will be set to this value in mutated containers
      --enable-debugging-handlers        Enable debugging handlers. Currently /debug/alpha/cache is supported
      --in-cluster                       Use in-cluster authentication and certificate request API (default true)
      --kube-api string                  (out-of-cluster) The url to the API server
      --kubeconfig string                (out-of-cluster) Absolute path to the API server kubeconfig file
      --log_backtrace_at traceLocation   when logging hits line file:N, emit a stack trace (default :0)
      --log_dir string                   If non-empty, write log files in this directory
      --log_file string                  If non-empty, use this log file
      --log_file_max_size uint           Defines the maximum size a log file can grow to. Unit is megabytes. If the value is 0, the maximum file size is unlimited. (default 1800)
      --logtostderr                      log to standard error instead of files (default true)
      --metrics-port int                 Port to listen on for metrics and healthz (http) (default 9999)
      --namespace string                 (in-cluster) The namespace name this webhook, the TLS secret, and configmap resides in (default "eks")
      --port int                         Port to listen on (default 443)
      --service-name string              (in-cluster) The service name fronting this webhook (default "pod-identity-webhook")
      --skip_headers                     If true, avoid header prefixes in the log messages
      --skip_log_headers                 If true, avoid headers when opening log files
      --stderrthreshold severity         logs at or above this threshold go to stderr (default 2)
      --sts-regional-endpoint false      Whether to inject the AWS_STS_REGIONAL_ENDPOINTS=regional env var in mutated pods. Defaults to false.
      --tls-cert string                  (out-of-cluster) TLS certificate file path (default "/etc/webhook/certs/tls.crt")
      --tls-key string                   (out-of-cluster) TLS key file path (default "/etc/webhook/certs/tls.key")
      --tls-secret string                (in-cluster) The secret name for storing the TLS serving cert (default "pod-identity-webhook")
      --token-audience string            The default audience for tokens. Can be overridden by annotation (default "")
      --token-expiration int             The token expiration (default 86400)
      --token-mount-path string          The path to mount tokens (default "/var/run/secrets/")
  -v, --v Level                          number for the log level verbosity
      --version                          Display the version and exit
      --vmodule moduleSpec               comma-separated list of pattern=N settings for file-filtered logging
      --watch-config-map                 Enables watching serviceaccounts that are configured through the pod-identity-webhook configmap instead of using annotations


When the aws-default-region flag is set this webhook will inject AWS_DEFAULT_REGION and AWS_REGION in mutated containers if AWS_DEFAULT_REGION and AWS_REGION are not already set.


When the sts-regional-endpoint flag is set to true, the webhook will inject the environment variable AWS_STS_REGIONAL_ENDPOINTS with the value set to regional. This environment variable will configure the AWS SDKs to perform the sts:AssumeRoleWithWebIdentity call to get credentials from the regional endpoint, instead of the global endpoint in us-east-1. This is desirable in almost all cases, unless the STS regional endpoint is disabled in your account.

You can also enable this per-service account with the annotation set to "true".

pod-identity-webhook ConfigMap

The purpose of the pod-identity-webhook ConfigMap is to simplify the mapping of IAM roles and ServiceAccount when using tools/installers like kOps that directly manage IAM roles and trust policies. When using these tools, users do not need to configure annotations on the ServiceAccounts as the tools already know the relationship can relay it to the webhook.

When the watch-config-map flag is set to true, the webhook will watch the pod-identity-webhook ConfigMap in the namespace configured by the --namespace flag for additional ServiceAccounts. The webhook will mutate Pods configured to use these ServiceAccounts even if they have no annotations.

Should the same ServiceAccount both be referenced both in the ConfigMap and have annotations, the annotations takes presedence.

Here is an example ConfigMap:

apiVersion: v1
  config: '{"default/myserviceaccount":{"RoleARN":"arn:aws-test:iam::123456789012:role/","Audience":"","UseRegionalSTS":true,"TokenExpiration":0},"myapp/myotherserviceaccount":{"RoleARN":"arn:aws-test:iam::123456789012:role/","Audience":"","UseRegionalSTS":true,"TokenExpiration":0},"test-*/myserviceaccount":{"RoleARN":"arn:aws-test:iam::123456789012:role/","Audience":"","UseRegionalSTS":true,"TokenExpiration":0}}'
kind: ConfigMap
  annotations: "443" https "true"
  creationTimestamp: null
  name: pod-identity-webhook
  namespace: kube-system

Container Images

Container images for amazon-eks-pod-identity-webhook can be found on Docker Hub.



You must install cert-manager as it is a pre-requisite for below deployments. (See cert-manager installation)


You can use the provided configuration files in the deploy directory, along with the provided Makefile.

make cluster-up IMAGE=amazon/amazon-eks-pod-identity-webhook:latest

This will:

  • Create a service account, role, cluster-role, role-binding, and cluster-role-binding that the deployment requires
  • Create the deployment, service, ClusterIssuer, certificate, and mutating webhook in the cluster
  • Use in-cluster=false so that the webhook reloads certificates from the filesystem rather than creating CSRs to request certificates (using CSRs is now deprecated and will not work versions later than v0.3.0).

For self-hosted API server configuration, see see

On API server



With the upgrade to client-go 1.18, certificate_manager_server_expiration_seconds metric has been removed by an upstream commit kubernetes/kubernetes#85874. A new metric certificate_manager_server_rotation_seconds is added which tracks the time a certificate was valid before getting rotated.

Code of Conduct



Apache 2.0 - Copyright 2019, Inc. or its affiliates. All Rights Reserved. See LICENSE