ITPub博客

首页 > 自动化运维 > DevOps > docker笔记29-k8s认证及serviceaccount、RBAC

docker笔记29-k8s认证及serviceaccount、RBAC

原创 DevOps 作者:czxin788 时间:2018-09-28 21:55:43 0 删除 编辑

    目前RBAC是k8s授权方式最常用的一种方式。

    在k8s上,一个客户端向apiserver发起请求,需要如下信息:

1)username,uid,
2) group,
3) extra(额外信息)
4) API
5) request path,例如:http://127.0.0.1:8080/apis/apps/v1/namespaces/kube-system/d
6)HTTP request action,如get,post,put,delete,
7)Http request action,如 get,list,create,udate,patch,watch,proxy,redirect,delete,deletecollection
8) Rresource
9)Subresource
10)Namespace
11)API group

    K8s可以支持多版本并存。

    其实,我们用kubectl向apiserver发起的命令,都是http方式的。

    k8s验证分为useraccount和serviceaccount。

    可以用代理:

[root@master ~]# kubectl proxy --port=8080
[root@master ~]# curl 
[root@master ~]# kubectl get deploy -n kube-system
NAME      DESIRED   CURRENT   UP-TO-DATE   AVAILABLE   AGE
coredns   2         2         2            2           20d
[root@master ~]# curl http://127.0.0.1:8080/apis/apps/v1/namespaces/kube-system/deployments


[root@master ~]# kubectl get svc
NAME         TYPE        CLUSTER-IP      EXTERNAL-IP   PORT(S)             AGE
kubernetes   ClusterIP   10.96.0.1       <none>        443/TCP             20d
[root@master ~]# kubectl describe svc kubernetes
Name:              kubernetes
Namespace:         default
Labels:            component=apiserver
                   provider=kubernetes
Annotations:       <none>
Selector:          <none>
Type:              ClusterIP
IP:                10.96.0.1
Port:              https  443/TCP
TargetPort:        6443/TCP
Endpoints:         172.16.1.100:6443
Session Affinity:  None
Events:            <none>

    上面我们看到10.96.0.1是kubernetes apiserver的地址,从而实现了集群外部通过10.96.0.1访问集群内部的pod,同时也实现了集群内部的pod访问集群外部的应用的功能。

    只要访问apiserver,就必须实现认证。而认证信息是存储在pod中的。

[root@master ~]# kubectl explain pods.spec.serviceAccountName
[root@master manifests]# kubectl create  serviceaccount  mysa -o yaml --dry-run > mysa.yaml
[root@master manifests]# cat mysa.yaml 
apiVersion: v1
kind: ServiceAccount
metadata:
  creationTimestamp: null
  name: mysa

    上面我们可以看到,只要是kubectl create的,只要加上-o yaml,就可以导出清单文件,这样我们以后就不用从头到尾写清单文件了,而是只要生产一个,然后改改就行了,这个很不错。

    另外kubectl get 也可以导出yaml格式的,如下:

[root@master manifests]# kubectl get pods myapp-1 -o yaml --export
apiVersion: v1
kind: Pod
metadata:
  creationTimestamp: null
  generateName: myapp-
  labels:
    app: myapp-pod
    controller-revision-hash: myapp-8598dd746f
    statefulset.kubernetes.io/pod-name: myapp-1
  ownerReferences:
  - apiVersion: apps/v1
    blockOwnerDeletion: true
    controller: true
    kind: StatefulSet
    name: myapp
    uid: a98ebc48-c24f-11e8-bb35-005056a24ecb
  selfLink: /api/v1/namespaces/default/pods/myapp-1
spec:
  containers:
  - image: ikubernetes/myapp:v1
    imagePullPolicy: IfNotPresent
    name: myapp
    ports:
    - containerPort: 80
      name: web
      protocol: TCP
    resources: {}
    terminationMessagePath: /dev/termination-log
    terminationMessagePolicy: File
    volumeMounts:
    - mountPath: /usr/share/nginx/html
      name: myappdata
    - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
      name: default-token-5r85r
      readOnly: true
  dnsPolicy: ClusterFirst
  hostname: myapp-1
  nodeName: node2
  priority: 0
  restartPolicy: Always
  schedulerName: default-scheduler
  securityContext: {}
  serviceAccount: default
  serviceAccountName: default
  subdomain: myapp-svc
  terminationGracePeriodSeconds: 30
  tolerations:
  - effect: NoExecute
    key: node.kubernetes.io/not-ready
    operator: Exists
    tolerationSeconds: 300
  - effect: NoExecute
    key: node.kubernetes.io/unreachable
    operator: Exists
    tolerationSeconds: 300
  volumes:
  - name: myappdata
    persistentVolumeClaim:
      claimName: myappdata-myapp-1
  - name: default-token-5r85r
    secret:
      defaultMode: 420
      secretName: default-token-5r85r
status:
  phase: Pending
  qosClass: BestEffort

    将上面的改改就成为我们新的配置清单了。

创建service account

[root@master manifests]# kubectl create serviceaccount admin
serviceaccount/admin created
[root@master manifests]# kubectl get sa
NAME      SECRETS   AGE
admin     1         2s
default   1         20d
[root@master manifests]# kubectl describe sa admin
Name:                admin
Namespace:           default
Labels:              <none>
Annotations:         <none>
Image pull secrets:  <none>
Mountable secrets:   admin-token-6jpc5
Tokens:              admin-token-6jpc5
Events:              <none>
[root@master manifests]# kubectl get secret
NAME                    TYPE                                  DATA      AGE
admin-token-6jpc5       kubernetes.io/service-account-token   3         57s

    看到自动就会多一个token。

    下面我们用配置清单把serviceaccount和pod绑定起来。

[root@master manifests]# cat  pod-sa-demo.yaml 
apiVersion: v1
kind: Pod
metadata:
  name: pod-sa-demo
  namespace: default
  labels:
    app: myapp
    tier: frontend
  annotations:
    zhixin.com/created-by: "cluster admin"
spec:
  containers:
  - name: myapp
    image: ikubernetes/myapp:v1
    ports:
    - name: http
      containerPort: 80
  serviceAccountName: admin #这表示我们这个pod使用自定义的验证信息admin
[root@master manifests]# kubectl apply -f pod-sa-demo.yaml 
pod/pod-sa-demo created

创建useraccount

    kubeconfig是客户端连接apiserver时使用的认证格式的配置文件。

[root@master manifests]# kubectl config view 
apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: REDACTED
    server: https://172.16.1.100:6443
  name: kubernetes
contexts:
- context: #context定义了哪个集群用哪个用户来访问。
    cluster: kubernetes
    user: kubernetes-admin
  name: kubernetes-admin@kubernetes
current-context: kubernetes-admin@kubernetes
kind: Config
preferences: {}
users:
- name: kubernetes-admin
  user:
    client-certificate-data: REDACTED
    client-key-data: REDACTED

    证书存放位置:

[root@master manifests]# cd /etc/kubernetes/pki/
[root@master pki]# ls
apiserver.crt              apiserver.key                 ca.crt  front-proxy-ca.crt      front-proxy-client.key
apiserver-etcd-client.crt  apiserver-kubelet-client.crt  ca.key  front-proxy-ca.key      sa.key
apiserver-etcd-client.key  apiserver-kubelet-client.key  etcd    front-proxy-client.crt  sa.pub

    例子:

     1、做一个私钥

[root@master pki]# cd /etc/kubernetes/pki
[root@master pki]# (umask 077; openssl genrsa -out zhixin.key 2048)
Generating RSA private key, 2048 bit long modulus
...........+++
...........+++
e is 65537 (0x10001)

    括号是子shell的意思。

     2、基于私钥生成一个证书

    CN就是用户的账户名字。

[root@master pki]# openssl req -new -key zhixin.key -out zhixin.csr -subj "/CN=zhixin"

  -subj:替换或指定证书申请者的个人信息


       3、签证书

[root@master pki]# openssl  x509 -req -in zhixin.csr -CA ca.crt  -CAkey ca.key  -CAcreateserial -out zhixin.crt -days 365
Signature ok
subject=/CN=zhixin
Getting CA Private Key

    -days:表示证书的过期时间

    x509:生成x509格式证书

     4、查看证书内容

[root@master pki]# openssl x509 -in zhixin.crt -text -noout
Certificate:
    Data:
        Version: 1 (0x0)
        Serial Number:
            ab:45:1b:b3:92:32:59:ae
    Signature Algorithm: sha256WithRSAEncryption
        Issuer: CN=kubernetes #证书签署人
        Validity #有效期限
            Not Before: Sep 28 08:01:20 2018 GMT
            Not After : Sep 28 08:01:20 2019 GMT
        Subject: CN=zhixin #一会用这个账户登录k8s
        Subject Public Key Info:
            Public Key Algorithm: rsaEncryption
                Public-Key: (2048 bit)
                Modulus:
                    00:bf:e5:b1:80:1a:a6:d1:24:ca:b8:75:a1:71:08:
                    d2:ba:43:ee:53:a1:10:b5:7a:83:e7:8b:06:65:c7:
                    8a:07:02:ca:cc:8f:5c:94:a9:7a:10:24:f6:41:a0:
                    c6:fe:5f:21:59:21:e7:72:30:12:38:89:85:78:54:
                    c1:15:c4:13:33:43:9c:94:c0:dc:99:e9:f0:44:7e:
                    35:66:cd:e0:d9:0c:82:dc:b3:73:ee:ea:47:9e:5e:
                    e5:bf:0b:45:fb:a3:cf:59:67:ae:13:31:9c:dc:b6:
                    78:da:b2:7e:c0:7e:c2:30:c5:fd:ea:6f:94:fa:81:
                    19:9f:71:9c:cf:60:07:5b:fa:0d:c0:6f:2c:b4:e0:
                    42:d6:6d:d3:39:23:2b:f7:ad:cc:21:f8:df:89:ff:
                    6e:45:59:1f:5d:db:aa:fa:07:ef:fc:b3:7e:3d:b1:
                    dd:3e:be:5e:43:de:8f:e2:ea:aa:ec:6c:48:df:2f:
                    2e:20:61:e3:5c:6a:37:3e:2b:32:e5:1a:ad:35:88:
                    d6:d2:db:aa:26:5d:cb:67:0a:65:9e:d4:79:76:92:
                    9a:41:fb:df:db:85:1a:ea:5e:ff:bb:7b:2f:01:10:
                    9f:8e:9c:a1:fe:ae:ac:9d:43:02:40:01:f7:d6:da:
                    bf:5a:99:ba:d0:bf:ea:53:1e:f5:51:06:9c:ac:6f:
                    32:43
                Exponent: 65537 (0x10001)
    Signature Algorithm: sha256WithRSAEncryption
         91:43:cd:36:ad:88:17:a1:81:9f:8f:ad:9b:c5:41:d7:de:aa:
         6a:f0:3a:00:f2:d7:9b:0e:89:bc:51:73:cc:4f:10:85:13:70:
         aa:d1:67:f8:f3:a1:6b:83:ff:99:76:7f:14:a5:b4:82:fb:1b:
         fb:cf:d5:fc:b0:2f:ff:68:c4:b1:c0:ee:f9:6b:41:ea:0a:96:
         2f:55:1d:d7:77:f8:70:a6:15:a4:b6:e7:6d:93:61:2e:ac:7a:
         10:70:fa:f7:43:da:56:f2:d0:e9:6b:01:72:73:2d:65:ea:4d:
         c4:3b:46:2d:1b:ad:f8:1f:eb:71:88:35:51:2a:dc:3a:36:fe:
         63:bb:28:ee:d2:a0:d4:e0:14:95:10:96:20:2e:f3:75:12:eb:
         05:8e:34:a1:dc:74:19:a5:76:0f:f2:bd:f3:56:aa:c9:40:51:
         c7:bd:1f:1f:c1:ec:a5:98:c8:b8:1d:07:67:fa:1c:a0:a3:1f:
         d3:ba:cb:09:52:9a:e7:59:39:ce:c8:ef:01:c2:4b:98:ff:05:
         12:bf:69:36:0e:a6:a9:f6:40:34:28:36:0d:1b:76:31:b4:96:
         6e:09:33:8e:d5:0a:96:77:dd:41:b3:29:db:d5:5e:fa:05:f7:
         e7:90:5d:79:6d:a9:59:20:60:0f:fe:d5:b6:38:6c:1a:ee:51:
         66:c3:9b:4b

     5、把用户账户信息添加到k8s集群中

[root@master pki]# kubectl config set-credentials zhixin --client-certificate=zhixin.crt --client-key=zhixin.key --embed-certs=true
User "zhixin" set.

embed-certs:表示把用户信息隐藏起来。

     5、设置context上下文,指定zhixin用户访问k8s的哪个集群

[root@master pki]# kubectl config view
apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: REDACTED
    server: https://172.16.1.100:6443
  name: kubernetes
contexts:
- context:
    cluster: kubernetes
    user: kubernetes-admin
  name: kubernetes-admin@kubernetes
current-context: kubernetes-admin@kubernetes
kind: Config
preferences: {}
users:
- name: kubernetes-admin
  user:
    client-certificate-data: REDACTED
    client-key-data: REDACTED
- name: zhixin
  user:
    client-certificate-data: REDACTED
    client-key-data: REDACTED
[root@master pki]# kubectl config set-context zhixin@kubernetes --cluster=kubernetes --user=zhixin
Context "zhixin@kubernetes" created.
[root@master pki]# kubectl config view
apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: REDACTED
    server: https://172.16.1.100:6443
  name: kubernetes
contexts:
- context:
    cluster: kubernetes
    user: kubernetes-admin
  name: kubernetes-admin@kubernetes
- context:
    cluster: kubernetes
    user: zhixin
  name: zhixin@kubernetes
current-context: kubernetes-admin@kubernetes
kind: Config
preferences: {}
users:
- name: kubernetes-admin
  user:
    client-certificate-data: REDACTED
    client-key-data: REDACTED
- name: zhixin
  user:
    client-certificate-data: REDACTED
    client-key-data: REDACTED

    上面看到contexts里面有zhixin的名字了。

    6、切换到zhixin用户登录k8s

[root@master pki]# kubectl config use-context zhixin@kubernetes
Switched to context "zhixin@kubernetes".
[root@master pki]# kubectl get pods
No resources found.
Error from server (Forbidden): pods is forbidden: User "zhixin" cannot list pods in the namespace "default"

      上面看到get pods时报错了,这是因为用户zhixin@kubernetes没有管理器权限。

     7、切回k8s管理员

[root@master pki]# kubectl config use-context kubernetes-admin@kubernetes
Switched to context "kubernetes-admin@kubernetes".

     8、设置新的k8s集群

[root@master ~]# kubectl config set-cluster mycluster --kubeconfig=/tmp/test.conf --server="https://127.0.0.1:6443" --certificate-authority=/etc/kubernetes/pki/ca.crt --embed-certs=true
Cluster "mycluster" set

.--kubeconfig:指定认证文件位置,不指定的话默认就在~/.kube/config 

--embed-certs=true 表示证书信息被隐藏

    大家看到,我们上面就创建了一个新的k8s集群叫mycluster。

[root@master ~]# kubectl config view --kubeconfig=/tmp/test.conf 
apiVersion: v1
clusters:
- cluster:
    certificate-authority-data: REDACTED
    server: https://127.0.0.1:6443
  name: mycluster
contexts: []
current-context: ""
kind: Config
preferences: {}
users: []

RBAC(基于角色的访问控制)

     rbac:role based ac,也就是我们把用户加入角色里面,这样用户就具有角色的权限了。

  

      

    在k8s中,一切皆对象。

    Object_URL: /apis/<GROUP>/<VERSION>/namespaces/<NAMESPACE_NAME>/<KIND>[OJJECT_ID]

     RBAC是通过rolebinding把user绑定到role上的。而role是基于namespace设定的,也就是这说这个user只能访问指定namespace下的pod资源。

     而如果把user通过clusterrolebind绑定到clusterrole上后,那么这个user就突破了namespace的限制,而拥有了集群级别的权限,即这个用户可以访问这个集群下所有namespace下的pod了。

     但是,我们也可以用rolebinding去把user绑定到clusterrole。在上图中,我们把user1通过rolebinding绑定到clusterrole上,但是我们知道rolebinding只限制在namespace中,所以user1也只限定在namespace中,而不是整个集群中。  

[root@master ~]# kubectl create role pods-reader --verb=get,list,watch --resource=pods
注意:想要授予所有权限可以用*来表示
[root@master ~]# kubectl create role pods-reader --verb=get,list,watch --resource=pods --dry-run -o yaml
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  creationTimestamp: null
  name: pods-reader
rules:
- apiGroups:
  - ""
  resources:
  - pods
  verbs:
  - get
  - list
  - watch
[root@master ~]# kubectl get role
NAME          AGE
pods-reader   7s
[root@master ~]# kubectl describe role pods-reader
Name:         pods-reader
Labels:       <none>
Annotations:  <none>
PolicyRule:
  Resources  Non-Resource URLs  Resource Names  Verbs
  ---------  -----------------  --------------  -----
  pods       []                 []              [get list watch]
[root@master ~]# kubectl create rolebinding zhixin-read-pods --role=pods-reader --user=zhixin
rolebinding.rbac.authorization.k8s.io/zhixin-read-pods created
[root@master ~]# kubectl create rolebinding zhixin-read-pods --role=pods-reader --user=zhixin -o yaml --dry-run
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  creationTimestamp: null
  name: zhixin-read-pods
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: pods-reader
subjects: #就是引用的用户
- apiGroup: rbac.authorization.k8s.io
  kind: User
  name: zhixin
[root@master ~]# kubectl explain rolebinding
[root@master ~]# kubectl describe rolebinding zhixin-read-pods
Name:         zhixin-read-pods
Labels:       <none>
Annotations:  <none>
Role:
  Kind:  Role
  Name:  pods-reader
Subjects:
  Kind  Name    Namespace
  ----  ----    ---------
  User  zhixin
[root@master ~]# kubectl config use-context zhixin@kubernetes
Switched to context "zhixin@kubernetes".
[root@master ~]# kubectl get pods
NAME                             READY     STATUS             RESTARTS   AGE
client                           0/1       Error              0          18d
filebeat-ds-bn7wf                0/1       InvalidImageName   0          4d
filebeat-ds-vd287                0/1       InvalidImageName   0          3d
liveness-httpget-pod             1/1       Running            7          11d
myapp-0                          1/1       Running            0          23h

    上面我们看到先前我们建立的zhixin用户是没有get pods权限的,但是我这回把它加入了pods-reader role,也就拥有了pods-reader role的权限。

[root@master ~]# kubectl get pods -n kube-system
No resources found.
Error from server (Forbidden): pods is forbidden: User "zhixin" cannot list pods in the namespace "kube-system

    但是,zhixin用户就没有访问kube-system权限,因为role就没有访问这个名称空间的权限,而只有访问default名称空间的权限。

    rolebinding只对namespace有效。

    我们再切换回到管理员。

[root@master ~]# kubectl config use-context kubernetes-admin@kubernetes
Switched to context "kubernetes-admin@kubernetes".

    下面我们再定义一个clusterrole。

[root@master ~]# kubectl create clusterrole cluster-reader --verb=get,list,watch --resource=pods
clusterrole.rbac.authorization.k8s.io/cluster-reader created
[root@master ~]# kubectl create clusterrole cluster-reader --verb=get,list,watch --resource=pods -o yaml --dry-run
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  creationTimestamp: null
  name: cluster-reader
rules:
- apiGroups:
  - ""
  resources:
  - pods
  verbs:
  - get
  - list
  - watch
[root@master ~]# kubectl explain clusterrole
[root@master ~]# kubectl get rolebinding
NAME               AGE
zhixin-read-pods   46m
[root@master ~]# kubectl delete rolebinding zhixin-read-pods
rolebinding.rbac.authorization.k8s.io "zhixin-read-pods" deleted

    下面我们测试把用户zhixin加入clusterrole里面:

[root@master ~]# kubectl create clusterrolebinding zhixin-read-all-pods --clusterrole=cluster-reader --user=zhixin 
clusterrolebinding.rbac.authorization.k8s.io/zhixin-read-all-pods created
[root@master ~]# kubectl get clusterrolebinding |grep read
zhixin-read-all-pods                                   2m
[root@master ~]# kubectl describe clusterrolebinding zhixin-read-all-pods
Name:         zhixin-read-all-pods
Labels:       <none>
Annotations:  <none>
Role:
  Kind:  ClusterRole
  Name:  cluster-reader
Subjects:
  Kind  Name    Namespace
  ----  ----    ---------
  User  zhixin
[root@master ~]# kubectl config use-context zhixin@kubernetes
Switched to context "zhixin@kubernetes".
[root@master ~]# kubectl get pods
NAME                             READY     STATUS             RESTARTS   AGE
client                           0/1       Error              0          18d
filebeat-ds-bn7wf                0/1       InvalidImageName   0          4d
[root@master ~]# kubectl get pods -n kube-system
NAME                             READY     STATUS    RESTARTS   AGE
coredns-78fcdf6894-2l2cf         1/1       Running   17         21d
coredns-78fcdf6894-dkkfq         1/1       Running   16         21d

    可见,我们把用户zhixin绑定到clusterrole后,这个 用户对所有的名称空间都有权限了。因为cluserrolebinding是针对集群的,而rolebinding是只针对namespace的。

    下面我们再测试一个,把用户用rolebinding绑定到cluserrole里面,看是什么效果:

[root@master ~]# kubectl config use-context kubernetes-admin@kubernetes
Switched to context "kubernetes-admin@kubernetes".
[root@master ~]# kubectl  delete clusterrolebinding zhixin-read-all-pods
clusterrolebinding.rbac.authorization.k8s.io "zhixin-read-all-pods" deleted
[root@master ~]# kubectl create rolebinding zhixin-read-pods --clusterrole=cluster-reader --user=zhixin 
rolebinding.rbac.authorization.k8s.io/zhixin-read-pods created
[root@master ~]# kubectl describe rolebinding zhixin-read-pods
Name:         zhixin-read-pods
Labels:       <none>
Annotations:  <none>
Role:
  Kind:  ClusterRole
  Name:  cluster-read
Subjects:
  Kind  Name    Namespace
  ----  ----    ---------
  User  zhixin
[root@master ~]# kubectl config use-context zhixin@kubernetes
Switched to context "zhixin@kubernetes".
[root@master ~]# kubectl get pods
NAME                             READY     STATUS             RESTARTS   AGE
client                           0/1       Error              0          18d
filebeat-ds-bn7wf                0/1       InvalidImageName   0          4d
filebeat-ds-vd287                0/1       InvalidImageName   0          3d
liveness-httpget-pod             1/1       Running            7          11d
[root@master ~]# kubectl get pods -n kube-system
No resources found.
Error from server (Forbidden): pods is forbidden: User "zhixin" cannot list pods in the namespace "kube-system"

    可以看出,clusterrole用rolebinding绑定后,会被降级到rolebinding所在的namespace里面。

[root@master ~]# kubectl get clusterrole admin -o yaml
 resources:
  - pods
  - pods/attach
  - pods/exec
  - pods/portforward
  - pods/proxy
  verbs:
  - create
  - delete
  - deletecollection
  - get
  - list
  - patch
  - update
  - watch
[root@master ~]# kubectl create rolebinding default-nameespace-admin --clusterrole=adin --user=zhixin
rolebinding.rbac.authorization.k8s.io/default-nameespace-admin created

    这样,我们就把zhixin设置为default名称空间的管理员,而不是其他名称空间的管理员。这就是用rolebinding绑定clusterrole的功能。

[root@master ~]# kubectl get clusterrolebinding cluster-admin -o yaml
- apiGroup: rbac.authorization.k8s.io
  kind: Group
  name: system:masters
[root@master pki]# openssl x509 -in ./apiserver-kubelet-client.crt -text -noout
        Subject: O=system:masters, CN=kube-apiserver-kubelet-client

    看到system:masters组具有管理员权限










来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/28916011/viewspace-2215100/,如需转载,请注明出处,否则将追究法律责任。

请登录后发表评论 登录
全部评论

注册时间:2014-06-03

  • 博文量
    178
  • 访问量
    504442