如何深刻理解 EKS IAM Authenticator 的实现机制

1、k8s Authentication

咱们知道 apiserver 有如下几种认证方式：

• X509 Client Certs
• Static Token File
• Bootstrap Tokens
• Static Password File
• Service Account Tokens
• OpenID Connect Tokens
• Webhook Token Authentication
• Authenticating Proxy

通常咱们使用二进制自建 k8s 集群，或者使用 kubeadm 建立的集群，集群的管理者，也就是 k8s 中 user 这个对象，都是使用的 X509 证书进行验证，若是属于同一个 CA 签发的，就承认这个用户；

企业为了减小用户管理的复杂程度，须要想办法如何把 IAM 的用户集成到 EKS 集群中，这样咱们就不须要耽误为 EKS 建立用户，减小一套用户的维护工做，那这是如何集成的呢，是经过了 k8s 的 Webhook Token Authentication，其认证过程架构图以下：

咱们能够从图中看到多了 Authentication 这一部分，经过 aws-iam-authenticator，咱们能够知道，这是运行在 EKS Control Plane 中的一组 DaemonSet Pod，用来接收 apiserver 的认证请求。

2、解剖 EKS

2.一、建立 EKS

咱们使用 eksctl 这个命令建立 EKS 集群，默认 eksctl 会调用 awscli 的 config，因此咱们须要先配置好 awscli，相关的用户或者角色具备建立 eks 集群的权限便可：

eksctl create cluster --name eks --region us-east-1 \
--node-type=t2.small --nodes 1 --ssh-public-key .ssh/id_rsa.pub \
--managed --zones us-east-1f,us-east-1c --vpc-nat-mode Disable

集群建立好以后，会自动帮咱们配置好 kubeclt 须要的配置文件，而且集群的建立者会自动获取集群 cluster-admin 的角色，拥有最高权限。

2.二、引入 aws-iam-authenticator

咱们首先能够经过 CloudWatch Logs 去查看 kube-apiserver 的启动参数，咱们能够看到有这样一组启动参数：

--authentication-token-webhook-config-file="/etc/kubernetes/authenticator/apiserver-webhook-kubeconfig.yaml"

说明咱们已经启动了 webhook 方式的认证，那参数后面的 yaml 文件里面是什么呢，咱们能够经过 aws-iam-authenticator 的 github 文档能够看到，咱们使用以下命令进行生成：

wangzan:~/k8s $ aws-iam-authenticator init -i `openssl rand 16 -hex`
INFO[2020-01-07T07:50:54Z] generated a new private key and certificate   certBytes=804 keyBytes=1192
INFO[2020-01-07T07:50:54Z] saving new key and certificate                certPath=cert.pem keyPath=key.pem
INFO[2020-01-07T07:50:54Z] loaded existing keypair                       certPath=cert.pem keyPath=key.pem
INFO[2020-01-07T07:50:54Z] writing webhook kubeconfig file               kubeconfigPath=aws-iam-authenticator.kubeconfig
INFO[2020-01-07T07:50:54Z] copy cert.pem to /var/aws-iam-authenticator/cert.pem on kubernetes master node(s) 
INFO[2020-01-07T07:50:54Z] copy key.pem to /var/aws-iam-authenticator/key.pem on kubernetes master node(s) 
INFO[2020-01-07T07:50:54Z] copy aws-iam-authenticator.kubeconfig to /etc/kubernetes/aws-iam-authenticator/kubeconfig.yaml on kubernetes master node(s) 
INFO[2020-01-07T07:50:54Z] configure your apiserver with `--authentication-token-webhook-config-file=/etc/kubernetes/aws-iam-authenticator/kubeconfig.yaml` to enable authentication with aws-iam-authenticator

查看生成的配置文件aws-iam-authenticator.kubeconfig

wangzan:~/k8s $ cat aws-iam-authenticator.kubeconfig 

# clusters refers to the remote service.
clusters:
  - name: aws-iam-authenticator
    cluster:
      certificate-authority-data: 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
      server: https://localhost:21362/authenticate
# users refers to the API Server's webhook configuration
# (we don't need to authenticate the API server).
users:
  - name: apiserver
# kubeconfig files require a context. Provide one for the API Server.
current-context: webhook
contexts:
- name: webhook
  context:
    cluster: aws-iam-authenticator
    user: apiserver

2.三、整个 IAM 认证过程

首先，咱们查看一下 kubectl 的配置文件信息：

apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: DATA+OMITTED
    server: https://93BEE997ED0F1C1BA3BD6C8395BE0756.sk1.us-east-1.eks.amazonaws.com
  name: eks.us-east-1.eksctl.io
contexts:
- context:
    cluster: eks.us-east-1.eksctl.io
    user: wangzan@eks.us-east-1.eksctl.io
  name: wangzan@eks.us-east-1.eksctl.io
current-context: wangzan@eks.us-east-1.eksctl.io
kind: Config
preferences: {}
users:
- name: wangzan@eks.us-east-1.eksctl.io
  user:
    exec:
      apiVersion: client.authentication.k8s.io/v1alpha1
      args:
      - token
      - -i
      - eks
      command: aws-iam-authenticator
      env: null

咱们能够看到 user 字段，这里不是使用证书进行认证的，而是使用的 aws-iam-authenticator client，其命令以下：

wangzan:~ $ aws-iam-authenticator token -i eks
{"kind":"ExecCredential","apiVersion":"client.authentication.k8s.io/v1alpha1","spec":{},"status":{"expirationTimestamp":"2020-01-07T08:23:23Z","token":"k8s-aws-v1.aHR0cHM6Ly9zdHMuYW1hem9uYXdzLmNvbS8_QWN0aW9uPUdldENhbGxlcklkZW50aXR5JlZlcnNpb249MjAxMS0wNi0xNSZYLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1DcmVkZW50aWFsPUFLSUE1TkFHSEY2TllYU01DTEhPJTJGMjAyMDAxMDclMkZ1cy1lYXN0LTElMkZzdHMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDIwMDEwN1QwODA5MjNaJlgtQW16LUV4cGlyZXM9MCZYLUFtei1TaWduZWRIZWFkZXJzPWhvc3QlM0J4LWs4cy1hd3MtaWQmWC1BbXotU2lnbmF0dXJlPTU2MjA5OTZhY2MzZGE3OWI3OGI0NDVjOTVkMTMyNmU0NjZmNTUyZTMzNDdkN2Y5MmExNGUwMzcwOTJiMzdmMDY"}}

这里实际上是向 sts 获取一个临时的 token，用做身份凭证，这个命令也等同于下面的命令：

wangzan:~ $ aws eks get-token --cluster-name eks 
{"status": {"expirationTimestamp": "2020-01-07T08:25:38Z", "token": "k8s-aws-v1.aHR0cHM6Ly9zdHMuYW1hem9uYXdzLmNvbS8_QWN0aW9uPUdldENhbGxlcklkZW50aXR5JlZlcnNpb249MjAxMS0wNi0xNSZYLUFtei1BbGdvcml0aG09QVdTNC1ITUFDLVNIQTI1NiZYLUFtei1FeHBpcmVzPTYwJlgtQW16LURhdGU9MjAyMDAxMDdUMDgxMTM4WiZYLUFtei1TaWduZWRIZWFkZXJzPWhvc3QlM0J4LWs4cy1hd3MtaWQmWC1BbXotU2VjdXJpdHktVG9rZW49JlgtQW16LUNyZWRlbnRpYWw9QUtJQTVOQUdIRjZOWVhTTUNMSE8lMkYyMDIwMDEwNyUyRnVzLWVhc3QtMSUyRnN0cyUyRmF3czRfcmVxdWVzdCZYLUFtei1TaWduYXR1cmU9NDUyYzA5ZTIwMzg2YjFmODU0NTU4YjhjNzBkNDA2MzdkYzM2Y2ExNzA5YWIxODQzNzE3NDdhY2IwYTUyNGIzYw"}, "kind": "ExecCredential", "spec": {}, "apiVersion": "client.authentication.k8s.io/v1alpha1"}

咱们回到上面的架构图，kubectl 会把这个获取的 token，放到 http 的请求头 Authorization 里面，发送给 apiserver，apiserver 收到以后，回去请求已经配置好的 webhook 服务器，也就是咱们前面所说的 daemonset pod（aws-iam-authenticator server），这个 aws-iam-authenticator server 会拿着 token 去请求 sts 服务，sts 服务验证其 token 的合法性，并返回 IAM 用户的 ARN（IAM Identity）；

当 aws-iam-authenticator server 获得返回的 ARN 以后，回去和 k8s 中的一个 configmap aws-auth 去对比。

2.四、configmap aws-auth

咱们查看一下刚建立好的机器，aws-auth 这个 configmap里面有什么信息：

wangzan:~ $ kubectl get cm aws-auth -nkube-system -oyaml
apiVersion: v1
data:
  mapRoles: |
    - groups:
      - system:bootstrappers
      - system:nodes
      rolearn: arn:aws:iam::921283538843:role/eksctl-eks-nodegroup-ng-5a1b33b9-NodeInstanceRole-1B757SI5DCABJ
      username: system:node:{{EC2PrivateDNSName}}
    - groups:
      - system:bootstrappers
      - system:nodes
      - system:node-proxier
      rolearn: arn:aws:iam::921283538843:role/eksctl-eks-cluster-FargatePodExecutionRole-DEAGGBFGQ9YB
      username: system:node:{{SessionName}}
kind: ConfigMap
metadata:
  creationTimestamp: "2019-12-30T07:57:47Z"
  name: aws-auth
  namespace: kube-system
  resourceVersion: "529891"
  selfLink: /api/v1/namespaces/kube-system/configmaps/aws-auth
  uid: 117c0e14-2ada-11ea-8820-0a64f353aa45

这里面定义了 IAM Identity 和 k8s 里面的 user 或者 group 的映射关系，集群建立的默认管理员并无放在这里面，多是处于安全的考虑，由于这个文件是能够进行编辑修改的。

经过上面一步咱们能够知道，经过对比 aws-auth，apiserver 会获得请求用户的 username 或者 group，而后在经过其 authorization 受权方式对 IAM 赋予相应的权限，咱们集群通常使用的是 RBAC。

关于更多的一些配置信息，能够参照官方文档：

https://github.com/kubernetes-sigs/aws-iam-authenticator
https://docs.aws.amazon.com/zh_cn/eks/latest/userguide/add-user-role.html