06-3.部署高可用 kube-controller-manager 集群
本文档介绍部署高可用 kube-controller-manager 集群的步骤。
该集群包含 3 个节点,启动后将通过竞争选举机制产生一个 leader 节点,其它节点为阻塞状态。当 leader 节点不可用后,剩余节点将再次进行选举产生新的 leader 节点,从而保证服务的可用性。
为保证通信安全,本文档先生成 x509 证书和私钥,kube-controller-manager 在如下两种情况下使用该证书:
- 与 kube-apiserver 的安全端口通信时;
- 在安全端口(https,10252) 输出 prometheus 格式的 metrics;
注意:如果没有特殊指明,本文档的所有操作均在 m7-autocv-gpu01 节点上执行,然后远程分发文件和执行命令。
准备工作
下载最新版本的二进制文件、安装和配置 flanneld 参考:06-1.部署master节点.md
创建 kube-controller-manager 证书和私钥
创建证书签名请求:
cd /opt/k8s/work
cat > kube-controller-manager-csr.json <<EOF
{
"CN": "system:kube-controller-manager",
"key": {
"algo": "rsa",
"size": 2048
},
"hosts": [
"127.0.0.1",
"172.27.128.150",
"172.27.128.149",
"172.27.128.148"
],
"names": [
{
"C": "CN",
"ST": "BeiJing",
"L": "BeiJing",
"O": "system:kube-controller-manager",
"OU": "4Paradigm"
}
]
}
EOF
- hosts 列表包含所有 kube-controller-manager 节点 IP;
- CN 为 system:kube-controller-manager、O 为 system:kube-controller-manager,kubernetes 内置的 ClusterRoleBindings system:kube-controller-manager 赋予 kube-controller-manager 工作所需的权限。
生成证书和私钥:
cd /opt/k8s/work
cfssl gencert -ca=/opt/k8s/work/ca.pem \
-ca-key=/opt/k8s/work/ca-key.pem \
-config=/opt/k8s/work/ca-config.json \
-profile=kubernetes kube-controller-manager-csr.json | cfssljson -bare kube-controller-manager
ls kube-controller-manager*pem
将生成的证书和私钥分发到所有 master 节点:
cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
do
echo ">>> ${node_ip}"
scp kube-controller-manager*.pem root@${node_ip}:/etc/kubernetes/cert/
done
创建和分发 kubeconfig 文件
kubeconfig 文件包含访问 apiserver 的所有信息,如 apiserver 地址、CA 证书和自身使用的证书;
cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
kubectl config set-cluster kubernetes \
--certificate-authority=/opt/k8s/work/ca.pem \
--embed-certs=true \
--server=${KUBE_APISERVER} \
--kubeconfig=kube-controller-manager.kubeconfig
kubectl config set-credentials system:kube-controller-manager \
--client-certificate=kube-controller-manager.pem \
--client-key=kube-controller-manager-key.pem \
--embed-certs=true \
--kubeconfig=kube-controller-manager.kubeconfig
kubectl config set-context system:kube-controller-manager \
--cluster=kubernetes \
--user=system:kube-controller-manager \
--kubeconfig=kube-controller-manager.kubeconfig
kubectl config use-context system:kube-controller-manager --kubeconfig=kube-controller-manager.kubeconfig
分发 kubeconfig 到所有 master 节点:
cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
do
echo ">>> ${node_ip}"
scp kube-controller-manager.kubeconfig root@${node_ip}:/etc/kubernetes/
done
创建和分发 kube-controller-manager systemd unit 文件
cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
cat > kube-controller-manager.service <<EOF
[Unit]
Description=Kubernetes Controller Manager
Documentation=https://github.com/GoogleCloudPlatform/kubernetes
[Service]
WorkingDirectory=${K8S_DIR}/kube-controller-manager
ExecStart=/opt/k8s/bin/kube-controller-manager \\
--port=0 \\
--secure-port=10252 \\
--bind-address=127.0.0.1 \\
--kubeconfig=/etc/kubernetes/kube-controller-manager.kubeconfig \\
--authentication-kubeconfig=/etc/kubernetes/kube-controller-manager.kubeconfig \\
--authorization-kubeconfig=/etc/kubernetes/kube-controller-manager.kubeconfig \\
--service-cluster-ip-range=${SERVICE_CIDR} \\
--cluster-name=kubernetes \\
--cluster-signing-cert-file=/etc/kubernetes/cert/ca.pem \\
--cluster-signing-key-file=/etc/kubernetes/cert/ca-key.pem \\
--experimental-cluster-signing-duration=8760h \\
--root-ca-file=/etc/kubernetes/cert/ca.pem \\
--service-account-private-key-file=/etc/kubernetes/cert/ca-key.pem \\
--leader-elect=true \\
--controllers=*,bootstrapsigner,tokencleaner \\
--horizontal-pod-autoscaler-use-rest-clients=true \\
--horizontal-pod-autoscaler-sync-period=10s \\
--tls-cert-file=/etc/kubernetes/cert/kube-controller-manager.pem \\
--tls-private-key-file=/etc/kubernetes/cert/kube-controller-manager-key.pem \\
--use-service-account-credentials=true \\
--kube-api-qps=1000 \\
--kube-api-burst=2000 \\
--logtostderr=true \\
--v=2
Restart=on-failure
RestartSec=5
[Install]
WantedBy=multi-user.target
EOF
--port=0
:关闭监听 http /metrics 的请求,同时--address
参数无效,--bind-address
参数有效;--secure-port=10252
、--bind-address=0.0.0.0
: 在所有网络接口监听 10252 端口的 https /metrics 请求;--kubeconfig
:指定 kubeconfig 文件路径,kube-controller-manager 使用它连接和验证 kube-apiserver;--authentication-kubeconfig
和--authorization-kubeconfig
:kube-controller-manager 使用它连接 apiserver,对 client 的请求进行认证和授权。kube-controller-manager
不再使用--tls-ca-file
对请求 https metrics 的 Client 证书进行校验。如果没有配置这两个 kubeconfig 参数,则 client 连接 kube-controller-manager https 端口的请求会被拒绝(提示权限不足)。--cluster-signing-*-file
:签名 TLS Bootstrap 创建的证书;--experimental-cluster-signing-duration
:指定 TLS Bootstrap 证书的有效期;--root-ca-file
:放置到容器 ServiceAccount 中的 CA 证书,用来对 kube-apiserver 的证书进行校验;--service-account-private-key-file
:签名 ServiceAccount 中 Token 的私钥文件,必须和 kube-apiserver 的--service-account-key-file
指定的公钥文件配对使用;--service-cluster-ip-range
:指定 Service Cluster IP 网段,必须和 kube-apiserver 中的同名参数一致;--leader-elect=true
:集群运行模式,启用选举功能;被选为 leader 的节点负责处理工作,其它节点为阻塞状态;--controllers=*,bootstrapsigner,tokencleaner
:启用的控制器列表,tokencleaner 用于自动清理过期的 Bootstrap token;--horizontal-pod-autoscaler-*
:custom metrics 相关参数,支持 autoscaling/v2alpha1;--tls-cert-file
、--tls-private-key-file
:使用 https 输出 metrics 时使用的 Server 证书和秘钥;--use-service-account-credentials=true
: kube-controller-manager 中各 controller 使用 serviceaccount 访问 kube-apiserver;
完整 unit 见 kube-controller-manager.service
分发 systemd unit 文件到所有 master 节点:
cd /opt/k8s/work
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
do
echo ">>> ${node_ip}"
scp kube-controller-manager.service root@${node_ip}:/etc/systemd/system/
done
kube-controller-manager 的权限
ClusteRole: system:kube-controller-manager 的权限很小,只能创建 secret、serviceaccount 等资源对象,各 controller 的权限分散到 ClusterRole system:controller:XXX 中。
需要在 kube-controller-manager 的启动参数中添加 --use-service-account-credentials=true
参数,这样 main controller 会为各 controller 创建对应的 ServiceAccount XXX-controller。
内置的 ClusterRoleBinding system:controller:XXX 将赋予各 XXX-controller ServiceAccount 对应的 ClusterRole system:controller:XXX 权限。
另外,--authentication-kubeconfig
和 --authorization-kubeconfig
参数指定的证书需要有创建 “subjectaccessreviews” 的权限,否则提示:
$ curl --cacert /opt/k8s/work/ca.pem --cert /opt/k8s/work/admin.pem --key /opt/k8s/work/admin-key.pem https://127.0.0.1:10252/metrics
Internal Server Error: "/metrics": subjectaccessreviews.authorization.k8s.io is forbidden: User "system:kube-controller-manager" cannot create resource "subjectaccessreviews" in API group "authorization.k8s.io" at the cluster scope
解决办法是创建一个 ClusterRoleBinding,赋予相应的权限:
$ kubectl create clusterrolebinding controller-manager:system:auth-delegator --user system:kube-controller-manager --clusterrole system:auth-delegator
clusterrolebinding.rbac.authorization.k8s.io/controller-manager:system:auth-delegator created
参考:https://github.com/kubernetes/kubeadm/issues/1285
启动 kube-controller-manager 服务
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
do
echo ">>> ${node_ip}"
ssh root@${node_ip} "mkdir -p ${K8S_DIR}/kube-controller-manager"
ssh root@${node_ip} "systemctl daemon-reload && systemctl enable kube-controller-manager && systemctl restart kube-controller-manager"
done
- 必须创建工作目录;
检查服务运行状态
source /opt/k8s/bin/environment.sh
for node_ip in ${NODE_IPS[@]}
do
echo ">>> ${node_ip}"
ssh root@${node_ip} "systemctl status kube-controller-manager|grep Active"
done
确保状态为 active (running)
,否则查看日志,确认原因:
$ journalctl -u kube-controller-manager
查看输出的 metric
注意:以下命令在 kube-controller-manager 节点上执行。
kube-controller-manager 监听 10252 端口,接收 https 请求:
$ sudo netstat -lnpt|grep kube-controll
tcp 0 0 127.0.0.1:10252 0.0.0.0:* LISTEN 55755/kube-controll
$ curl --cacert /opt/k8s/work/ca.pem --cert /opt/k8s/work/admin.pem --key /opt/k8s/work/admin-key.pem https://127.0.0.1:10252/metrics
{
"kind": "Status",
"apiVersion": "v1",
"metadata": {
},
"status": "Failure",
"message": "forbidden: User \"system:anonymous\" cannot get path \"/metrics\"",
"reason": "Forbidden",
"details": {
},
"code": 403
}
将 kube-controller-manager 的日志级别设置为 4 后,可以看到原因是:
$ journalctl -u kube-controller-manager -f |grep /metrics
2月 22 19:07:28 m7-inf-prod01 kube-controller-manager[1416748]: I0222 19:07:28.003325 1416748 authorization.go:73] Forbidden: "/metrics", Reason: "no RBAC policy matched"
2月 22 19:07:28 m7-inf-prod01 kube-controller-manager[1416748]: I0222 19:07:28.003472 1416748 wrap.go:42] GET /metrics: (2.600519ms) 403 [curl/7.29.0 127.0.0.1:36324]
这是由于没有部署 metrics-server 的缘故。后续在 09-4.metrics-server插件.md 一文中将介绍部署 metrics-server 的步骤。
参考:https://github.com/kubernetes-incubator/metrics-server/issues/85
测试 kube-controller-manager 集群的高可用
停掉一个或两个节点的 kube-controller-manager 服务,观察其它节点的日志,看是否获取了 leader 权限。
查看当前的 leader
$ kubectl get endpoints kube-controller-manager --namespace=kube-system -o yaml
apiVersion: v1
kind: Endpoints
metadata:
annotations:
control-plane.alpha.kubernetes.io/leader: '{"holderIdentity":"m7-autocv-gpu02_084534e2-6cc4-11e8-a418-5254001f5b65","leaseDurationSeconds":15,"acquireTime":"2018-06-10T15:40:33Z","renewTime":"2018-06-10T16:19:08Z","leaderTransitions":12}'
creationTimestamp: 2018-06-10T13:59:42Z
name: kube-controller-manager
namespace: kube-system
resourceVersion: "4540"
selfLink: /api/v1/namespaces/kube-system/endpoints/kube-controller-manager
uid: 862cc048-6cb6-11e8-96fa-525400ba84c6
可见,当前的 leader 为 m7-autocv-gpu02 节点。
参考
- 关于 controller 权限和 use-service-account-credentials 参数:https://github.com/kubernetes/kubernetes/issues/48208
- kubelet 认证和授权:https://kubernetes.io/docs/admin/kubelet-authentication-authorization/#kubelet-authorization