06-4. 部署 kubelet 组件

06-4. 部署 kubelet 组件

kubelet 运行在每个 worker 节点上，接收 kube-apiserver 发送的请求，管理 Pod 容器，执行交互式命令，如 exec、run、logs 等。

kubelet 启动时自动向 kube-apiserver 注册节点信息，内置的 cadvisor 统计和监控节点的资源使用情况。

为确保安全，部署时关闭了 kubelet 的非安全 http 端口，对请求进行认证和授权，拒绝未授权的访问(如 apiserver、heapster 的请求)。

注意：如果没有特殊指明，本文档的所有操作均在 zhangjun-k8s-01 节点上执行。

下载和分发 kubelet 二进制文件

创建 kubelet bootstrap kubeconfig 文件

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_name in ${NODE_NAMES[@]}
  do
    echo ">>> ${node_name}"
    # 创建 token
    export BOOTSTRAP_TOKEN=$(kubeadm token create \
      --description kubelet-bootstrap-token \
      --groups system:bootstrappers:${node_name} \
      --kubeconfig ~/.kube/config)
    # 设置集群参数
    kubectl config set-cluster kubernetes \
      --certificate-authority=/etc/kubernetes/cert/ca.pem \
      --embed-certs=true \
      --server=${KUBE_APISERVER} \
      --kubeconfig=kubelet-bootstrap-${node_name}.kubeconfig
    # 设置客户端认证参数
    kubectl config set-credentials kubelet-bootstrap \
      --token=${BOOTSTRAP_TOKEN} \
      --kubeconfig=kubelet-bootstrap-${node_name}.kubeconfig
    # 设置上下文参数
    kubectl config set-context default \
      --cluster=kubernetes \
      --user=kubelet-bootstrap \
      --kubeconfig=kubelet-bootstrap-${node_name}.kubeconfig
    # 设置默认上下文
    kubectl config use-context default --kubeconfig=kubelet-bootstrap-${node_name}.kubeconfig
  done

向 kubeconfig 写入的是 token，bootstrap 结束后 kube-controller-manager 为 kubelet 创建 client 和 server 证书；

查看 kubeadm 为各节点创建的 token：

$ kubeadm token list --kubeconfig ~/.kube/config
TOKEN                     TTL       EXPIRES                     USAGES                   DESCRIPTION               EXTRA GROUPS
2sb8wy.euialqfpxfbcljby   23h       2020-02-08T15:36:30+08:00   authentication,signing   kubelet-bootstrap-token   system:bootstrappers:zhangjun-k8s-02
ta7onm.fcen74h0mczyfbz2   23h       2020-02-08T15:36:30+08:00   authentication,signing   kubelet-bootstrap-token   system:bootstrappers:zhangjun-k8s-01
xk27zp.tylnvywx9kc8sq87   23h       2020-02-08T15:36:30+08:00   authentication,signing   kubelet-bootstrap-token   system:bootstrappers:zhangjun-k8s-03

token 有效期为 1 天，超期后将不能再被用来 boostrap kubelet，且会被 kube-controller-manager 的 tokencleaner 清理；
kube-apiserver 接收 kubelet 的 bootstrap token 后，将请求的 user 设置为 system:bootstrap:<Token ID>，group 设置为 system:bootstrappers，后续将为这个 group 设置 ClusterRoleBinding；

分发 bootstrap kubeconfig 文件到所有 worker 节点

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_name in ${NODE_NAMES[@]}
  do
    echo ">>> ${node_name}"
    scp kubelet-bootstrap-${node_name}.kubeconfig root@${node_name}:/etc/kubernetes/kubelet-bootstrap.kubeconfig
  done

创建和分发 kubelet 参数配置文件

从 v1.10 开始，部分 kubelet 参数需在配置文件中配置，kubelet --help 会提示：

DEPRECATED: This parameter should be set via the config file specified by the Kubelet's --config flag

创建 kubelet 参数配置文件模板（可配置项参考代码中注释）：

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
cat > kubelet-config.yaml.template <<EOF
kind: KubeletConfiguration
apiVersion: kubelet.config.k8s.io/v1beta1
address: "##NODE_IP##"
staticPodPath: ""
syncFrequency: 1m
fileCheckFrequency: 20s
httpCheckFrequency: 20s
staticPodURL: ""
port: 10250
readOnlyPort: 0
rotateCertificates: true
serverTLSBootstrap: true
authentication:
  anonymous:
    enabled: false
  webhook:
    enabled: true
  x509:
    clientCAFile: "/etc/kubernetes/cert/ca.pem"
authorization:
  mode: Webhook
registryPullQPS: 0
registryBurst: 20
eventRecordQPS: 0
eventBurst: 20
enableDebuggingHandlers: true
enableContentionProfiling: true
healthzPort: 10248
healthzBindAddress: "##NODE_IP##"
clusterDomain: "${CLUSTER_DNS_DOMAIN}"
clusterDNS:
  - "${CLUSTER_DNS_SVC_IP}"
nodeStatusUpdateFrequency: 10s
nodeStatusReportFrequency: 1m
imageMinimumGCAge: 2m
imageGCHighThresholdPercent: 85
imageGCLowThresholdPercent: 80
volumeStatsAggPeriod: 1m
kubeletCgroups: ""
systemCgroups: ""
cgroupRoot: ""
cgroupsPerQOS: true
cgroupDriver: cgroupfs
runtimeRequestTimeout: 10m
hairpinMode: promiscuous-bridge
maxPods: 220
podCIDR: "${CLUSTER_CIDR}"
podPidsLimit: -1
resolvConf: /etc/resolv.conf
maxOpenFiles: 1000000
kubeAPIQPS: 1000
kubeAPIBurst: 2000
serializeImagePulls: false
evictionHard:
  memory.available:  "100Mi"
  nodefs.available:  "10%"
  nodefs.inodesFree: "5%"
  imagefs.available: "15%"
evictionSoft: {}
enableControllerAttachDetach: true
failSwapOn: true
containerLogMaxSize: 20Mi
containerLogMaxFiles: 10
systemReserved: {}
kubeReserved: {}
systemReservedCgroup: ""
kubeReservedCgroup: ""
enforceNodeAllocatable: ["pods"]
EOF

address：kubelet 安全端口（https，10250）监听的地址，不能为 127.0.0.1，否则 kube-apiserver、heapster 等不能调用 kubelet 的 API；
readOnlyPort=0：关闭只读端口(默认 10255)，等效为未指定；
authentication.anonymous.enabled：设置为 false，不允许匿名访问 10250 端口；
authentication.x509.clientCAFile：指定签名客户端证书的 CA 证书，开启 HTTP 证书认证；
authentication.webhook.enabled=true：开启 HTTPs bearer token 认证；
对于未通过 x509 证书和 webhook 认证的请求(kube-apiserver 或其他客户端)，将被拒绝，提示 Unauthorized；
authroization.mode=Webhook：kubelet 使用 SubjectAccessReview API 查询 kube-apiserver 某 user、group 是否具有操作资源的权限(RBAC)；
featureGates.RotateKubeletClientCertificate、featureGates.RotateKubeletServerCertificate：自动 rotate 证书，证书的有效期取决于 kube-controller-manager 的 —experimental-cluster-signing-duration 参数；
需要 root 账户运行；

为各节点创建和分发 kubelet 配置文件：

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
  do 
    echo ">>> ${node_ip}"
    sed -e "s/##NODE_IP##/${node_ip}/" kubelet-config.yaml.template > kubelet-config-${node_ip}.yaml.template
    scp kubelet-config-${node_ip}.yaml.template root@${node_ip}:/etc/kubernetes/kubelet-config.yaml
  done

创建和分发 kubelet systemd unit 文件

创建 kubelet systemd unit 文件模板：

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
cat > kubelet.service.template <<EOF
[Unit]
Description=Kubernetes Kubelet
Documentation=https://github.com/GoogleCloudPlatform/kubernetes
After=containerd.service
Requires=containerd.service
[Service]
WorkingDirectory=${K8S_DIR}/kubelet
ExecStart=/opt/k8s/bin/kubelet \\
  --bootstrap-kubeconfig=/etc/kubernetes/kubelet-bootstrap.kubeconfig \\
  --cert-dir=/etc/kubernetes/cert \\
  --network-plugin=cni \\
  --cni-conf-dir=/etc/cni/net.d \\
  --container-runtime=remote \\
  --container-runtime-endpoint=unix:///var/run/containerd/containerd.sock \\
  --root-dir=${K8S_DIR}/kubelet \\
  --kubeconfig=/etc/kubernetes/kubelet.kubeconfig \\
  --config=/etc/kubernetes/kubelet-config.yaml \\
  --hostname-override=##NODE_NAME## \\
  --image-pull-progress-deadline=15m \\
  --volume-plugin-dir=${K8S_DIR}/kubelet/kubelet-plugins/volume/exec/ \\
  --logtostderr=true \\
  --v=2
Restart=always
RestartSec=5
StartLimitInterval=0
[Install]
WantedBy=multi-user.target
EOF

如果设置了 --hostname-override 选项，则 kube-proxy 也需要设置该选项，否则会出现找不到 Node 的情况；
--bootstrap-kubeconfig：指向 bootstrap kubeconfig 文件，kubelet 使用该文件中的用户名和 token 向 kube-apiserver 发送 TLS Bootstrapping 请求；
K8S approve kubelet 的 csr 请求后，在 --cert-dir 目录创建证书和私钥文件，然后写入 --kubeconfig 文件；
--pod-infra-container-image 不使用 redhat 的 pod-infrastructure:latest 镜像，它不能回收容器的僵尸；

为各节点创建和分发 kubelet systemd unit 文件：

cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_name in ${NODE_NAMES[@]}
  do 
    echo ">>> ${node_name}"
    sed -e "s/##NODE_NAME##/${node_name}/" kubelet.service.template > kubelet-${node_name}.service
    scp kubelet-${node_name}.service root@${node_name}:/etc/systemd/system/kubelet.service
  done

授予 kube-apiserver 访问 kubelet API 的权限

在执行 kubectl exec、run、logs 等命令时，apiserver 会将请求转发到 kubelet 的 https 端口。这里定义 RBAC 规则，授权 apiserver 使用的证书（kubernetes.pem）用户名（CN：kuberntes-master）访问 kubelet API 的权限：

kubectl create clusterrolebinding kube-apiserver:kubelet-apis --clusterrole=system:kubelet-api-admin --user kubernetes-master

Bootstrap Token Auth 和授予权限

kubelet 启动时查找 --kubeletconfig 参数对应的文件是否存在，如果不存在则使用 --bootstrap-kubeconfig 指定的 kubeconfig 文件向 kube-apiserver 发送证书签名请求 (CSR)。

kube-apiserver 收到 CSR 请求后，对其中的 Token 进行认证，认证通过后将请求的 user 设置为 system:bootstrap:<Token ID>，group 设置为 system:bootstrappers，这一过程称为 Bootstrap Token Auth。

默认情况下，这个 user 和 group 没有创建 CSR 的权限，kubelet 启动失败，错误日志如下：

$ sudo journalctl -u kubelet -a |grep -A 2 'certificatesigningrequests'
May 26 12:13:41 zhangjun-k8s-01 kubelet[128468]: I0526 12:13:41.798230  128468 certificate_manager.go:366] Rotating certificates
May 26 12:13:41 zhangjun-k8s-01 kubelet[128468]: E0526 12:13:41.801997  128468 certificate_manager.go:385] Failed while requesting a signed certificate from the master: cannot create certificate signing request: certificatesigningrequests.certificates.k8s.io is forbidden: User "system:bootstrap:82jfrm" cannot create resource "certificatesigningrequests" in API group "certificates.k8s.io" at the cluster scope

解决办法是：创建一个 clusterrolebinding，将 group system:bootstrappers 和 clusterrole system:node-bootstrapper 绑定：

kubectl create clusterrolebinding kubelet-bootstrap --clusterrole=system:node-bootstrapper --group=system:bootstrappers

自动 approve CSR 请求，生成 kubelet client 证书

kubelet 创建 CSR 请求后，下一步需要创建被 approve，有两种方式：

kube-controller-manager 自动 aprrove；
手动使用命令 kubectl certificate approve；

CSR 被 approve 后，kubelet 向 kube-controller-manager 请求创建 client 证书，kube-controller-manager 中的 csrapproving controller 使用 SubjectAccessReview API 来检查 kubelet 请求（对应的 group 是 system:bootstrappers）是否具有相应的权限。

创建三个 ClusterRoleBinding，分别授予 group system:bootstrappers 和 group system:nodes 进行 approve client、renew client、renew server 证书的权限（server csr 是手动 approve 的，见后文）：

cd /opt/k8s/work
cat > csr-crb.yaml <<EOF
 # Approve all CSRs for the group "system:bootstrappers"
 kind: ClusterRoleBinding
 apiVersion: rbac.authorization.k8s.io/v1
 metadata:
   name: auto-approve-csrs-for-group
 subjects:
 - kind: Group
   name: system:bootstrappers
   apiGroup: rbac.authorization.k8s.io
 roleRef:
   kind: ClusterRole
   name: system:certificates.k8s.io:certificatesigningrequests:nodeclient
   apiGroup: rbac.authorization.k8s.io
---
 # To let a node of the group "system:nodes" renew its own credentials
 kind: ClusterRoleBinding
 apiVersion: rbac.authorization.k8s.io/v1
 metadata:
   name: node-client-cert-renewal
 subjects:
 - kind: Group
   name: system:nodes
   apiGroup: rbac.authorization.k8s.io
 roleRef:
   kind: ClusterRole
   name: system:certificates.k8s.io:certificatesigningrequests:selfnodeclient
   apiGroup: rbac.authorization.k8s.io
---
# A ClusterRole which instructs the CSR approver to approve a node requesting a
# serving cert matching its client cert.
kind: ClusterRole
apiVersion: rbac.authorization.k8s.io/v1
metadata:
  name: approve-node-server-renewal-csr
rules:
- apiGroups: ["certificates.k8s.io"]
  resources: ["certificatesigningrequests/selfnodeserver"]
  verbs: ["create"]
---
 # To let a node of the group "system:nodes" renew its own server credentials
 kind: ClusterRoleBinding
 apiVersion: rbac.authorization.k8s.io/v1
 metadata:
   name: node-server-cert-renewal
 subjects:
 - kind: Group
   name: system:nodes
   apiGroup: rbac.authorization.k8s.io
 roleRef:
   kind: ClusterRole
   name: approve-node-server-renewal-csr
   apiGroup: rbac.authorization.k8s.io
EOF
kubectl apply -f csr-crb.yaml

auto-approve-csrs-for-group：自动 approve node 的第一次 CSR；注意第一次 CSR 时，请求的 Group 为 system:bootstrappers；
node-client-cert-renewal：自动 approve node 后续过期的 client 证书，自动生成的证书 Group 为 system:nodes;
node-server-cert-renewal：自动 approve node 后续过期的 server 证书，自动生成的证书 Group 为 system:nodes;

启动 kubelet 服务

source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
  do
    echo ">>> ${node_ip}"
    ssh root@${node_ip} "mkdir -p ${K8S_DIR}/kubelet/kubelet-plugins/volume/exec/"
    ssh root@${node_ip} "/usr/sbin/swapoff -a"
    ssh root@${node_ip} "systemctl daemon-reload && systemctl enable kubelet && systemctl restart kubelet"
  done

启动服务前必须先创建工作目录；
关闭 swap 分区，否则 kubelet 会启动失败；

kubelet 启动后使用 —bootstrap-kubeconfig 向 kube-apiserver 发送 CSR 请求，当这个 CSR 被 approve 后，kube-controller-manager 为 kubelet 创建 TLS 客户端证书、私钥和 —kubeletconfig 文件。

注意：kube-controller-manager 需要配置 --cluster-signing-cert-file 和 --cluster-signing-key-file 参数，才会为 TLS Bootstrap 创建证书和私钥。

查看 kubelet 情况

稍等一会，三个节点的 CSR 都被自动 approved：

$ kubectl get csr
NAME        AGE   REQUESTOR                     CONDITION
csr-5rwzm   43s   system:node:zhangjun-k8s-01   Pending
csr-65nms   55s   system:bootstrap:2sb8wy       Approved,Issued
csr-8t5hj   42s   system:node:zhangjun-k8s-02   Pending
csr-jkhhs   41s   system:node:zhangjun-k8s-03   Pending
csr-jv7dn   56s   system:bootstrap:ta7onm       Approved,Issued
csr-vb6p5   54s   system:bootstrap:xk27zp       Approved,Issued

Pending 的 CSR 用于创建 kubelet server 证书，需要手动 approve，参考后文。

所有节点均注册（NotReady 状态是预期的，后续安装了网络插件后就好）：

$ kubectl get node
NAME              STATUS     ROLES    AGE   VERSION
zhangjun-k8s-01   NotReady   <none>   10h   v1.16.6
zhangjun-k8s-02   NotReady   <none>   10h   v1.16.6
zhangjun-k8s-03   NotReady   <none>   10h   v1.16.6

kube-controller-manager 为各 node 生成了 kubeconfig 文件和公私钥：

$ ls -l /etc/kubernetes/kubelet.kubeconfig
-rw------- 1 root root 2246 Feb  7 15:38 /etc/kubernetes/kubelet.kubeconfig
$ ls -l /etc/kubernetes/cert/kubelet-client-*
-rw------- 1 root root 1281 Feb  7 15:38 /etc/kubernetes/cert/kubelet-client-2020-02-07-15-38-21.pem
lrwxrwxrwx 1 root root   59 Feb  7 15:38 /etc/kubernetes/cert/kubelet-client-current.pem -> /etc/kubernetes/cert/kubelet-client-2020-02-07-15-38-21.pem

没有自动生成 kubelet server 证书；

手动 approve server cert csr

基于安全性考虑，CSR approving controllers 不会自动 approve kubelet server 证书签名请求，需要手动 approve：

$ kubectl get csr
NAME        AGE     REQUESTOR                     CONDITION
csr-5rwzm   3m22s   system:node:zhangjun-k8s-01   Pending
csr-65nms   3m34s   system:bootstrap:2sb8wy       Approved,Issued
csr-8t5hj   3m21s   system:node:zhangjun-k8s-02   Pending
csr-jkhhs   3m20s   system:node:zhangjun-k8s-03   Pending
csr-jv7dn   3m35s   system:bootstrap:ta7onm       Approved,Issued
csr-vb6p5   3m33s   system:bootstrap:xk27zp       Approved,Issued
$ # 手动 approve
$ kubectl get csr | grep Pending | awk '{print $1}' | xargs kubectl certificate approve
$ # 自动生成了 server 证书
$  ls -l /etc/kubernetes/cert/kubelet-*
-rw------- 1 root root 1281 Feb  7 15:38 /etc/kubernetes/cert/kubelet-client-2020-02-07-15-38-21.pem
lrwxrwxrwx 1 root root   59 Feb  7 15:38 /etc/kubernetes/cert/kubelet-client-current.pem -> /etc/kubernetes/cert/kubelet-client-2020-02-07-15-38-21.pem
-rw------- 1 root root 1330 Feb  7 15:42 /etc/kubernetes/cert/kubelet-server-2020-02-07-15-42-12.pem
lrwxrwxrwx 1 root root   59 Feb  7 15:42 /etc/kubernetes/cert/kubelet-server-current.pem -> /etc/kubernetes/cert/kubelet-server-2020-02-07-15-42-12.pem

kubelet api 认证和授权

kubelet 配置了如下认证参数：

authentication.anonymous.enabled：设置为 false，不允许匿名访问 10250 端口；
authentication.x509.clientCAFile：指定签名客户端证书的 CA 证书，开启 HTTPs 证书认证；
authentication.webhook.enabled=true：开启 HTTPs bearer token 认证；

同时配置了如下授权参数：

authroization.mode=Webhook：开启 RBAC 授权；

kubelet 收到请求后，使用 clientCAFile 对证书签名进行认证，或者查询 bearer token 是否有效。如果两者都没通过，则拒绝请求，提示 Unauthorized：

$ curl -s --cacert /etc/kubernetes/cert/ca.pem https://172.27.138.251:10250/metrics
Unauthorized
$ curl -s --cacert /etc/kubernetes/cert/ca.pem -H "Authorization: Bearer 123456" https://172.27.138.251:10250/metrics
Unauthorized

通过认证后，kubelet 使用 SubjectAccessReview API 向 kube-apiserver 发送请求，查询证书或 token 对应的 user、group 是否有操作资源的权限(RBAC)；

证书认证和授权

$ # 权限不足的证书；
$ curl -s --cacert /etc/kubernetes/cert/ca.pem --cert /etc/kubernetes/cert/kube-controller-manager.pem --key /etc/kubernetes/cert/kube-controller-manager-key.pem https://172.27.138.251:10250/metrics
Forbidden (user=system:kube-controller-manager, verb=get, resource=nodes, subresource=metrics)
$ # 使用部署 kubectl 命令行工具时创建的、具有最高权限的 admin 证书；
$ curl -s --cacert /etc/kubernetes/cert/ca.pem --cert /opt/k8s/work/admin.pem --key /opt/k8s/work/admin-key.pem https://172.27.138.251:10250/metrics|head
# HELP apiserver_audit_event_total Counter of audit events generated and sent to the audit backend.
# TYPE apiserver_audit_event_total counter
apiserver_audit_event_total 0
# HELP apiserver_audit_requests_rejected_total Counter of apiserver requests rejected due to an error in audit logging backend.
# TYPE apiserver_audit_requests_rejected_total counter
apiserver_audit_requests_rejected_total 0
# HELP apiserver_client_certificate_expiration_seconds Distribution of the remaining lifetime on the certificate used to authenticate a request.
# TYPE apiserver_client_certificate_expiration_seconds histogram
apiserver_client_certificate_expiration_seconds_bucket{le="0"} 0
apiserver_client_certificate_expiration_seconds_bucket{le="1800"} 0

--cacert、--cert、--key 的参数值必须是文件路径，如上面的 ./admin.pem 不能省略 ./，否则返回 401 Unauthorized；

bear token 认证和授权

创建一个 ServiceAccount，将它和 ClusterRole system:kubelet-api-admin 绑定，从而具有调用 kubelet API 的权限：

kubectl create sa kubelet-api-test
kubectl create clusterrolebinding kubelet-api-test --clusterrole=system:kubelet-api-admin --serviceaccount=default:kubelet-api-test
SECRET=$(kubectl get secrets | grep kubelet-api-test | awk '{print $1}')
TOKEN=$(kubectl describe secret ${SECRET} | grep -E '^token' | awk '{print $2}')
echo ${TOKEN}

$ curl -s --cacert /etc/kubernetes/cert/ca.pem -H "Authorization: Bearer ${TOKEN}" https://172.27.138.251:10250/metrics | head
# HELP apiserver_audit_event_total Counter of audit events generated and sent to the audit backend.
# TYPE apiserver_audit_event_total counter
apiserver_audit_event_total 0
# HELP apiserver_audit_requests_rejected_total Counter of apiserver requests rejected due to an error in audit logging backend.
# TYPE apiserver_audit_requests_rejected_total counter
apiserver_audit_requests_rejected_total 0
# HELP apiserver_client_certificate_expiration_seconds Distribution of the remaining lifetime on the certificate used to authenticate a request.
# TYPE apiserver_client_certificate_expiration_seconds histogram
apiserver_client_certificate_expiration_seconds_bucket{le="0"} 0
apiserver_client_certificate_expiration_seconds_bucket{le="1800"} 0