騰訊雲tke容器服務調整kubelet工做目錄

騰訊雲默認kubelet工做目錄在/var/lib/kubelet使用的是系統盤，系統盤通常不夠大也不方便擴容，因此須要修改kubelet工做目錄。大概流程以下：

• 準備工做

• 驅逐節點

• 修改配置

• 重啓服務

• 解封節點

• 驗證服務

準備工做
查找kubelet啓動文件，通常配置信息在啓動文件中

[root@VM_8_9_centos ~]# locate kubelet.service
/etc/systemd/system/multi-user.target.wants/kubelet.service
/usr/lib/systemd/system/kubelet.service
[root@VM_8_9_centos ~]# ls -lh /usr/lib/systemd/system/kubelet.service 
[root@VM_8_9_centos ~]# ls -lh  /etc/systemd/system/multi-user.target.wants/kubelet.service
lrwxrwxrwx 1 root root 39 Sep 25 11:59 /etc/systemd/system/multi-user.target.wants/kubelet.service -> /usr/lib/systemd/system/kubelet.service

[root@VM_8_9_centos ~]# cat /usr/lib/systemd/system/kubelet.service 

[Unit]
Description=kubelet

[Service]
Environment=QCLOUD_NORM_URL=
EnvironmentFile=-/etc/kubernetes/kubelet
ExecStart=/usr/bin/kubelet ${CNI_BIN_DIR} ${KUBE_RESERVED} ${MAX_PODS} ${CLOUD_PROVIDER} ${CLOUD_CONFIG} ${POD_INFRA_CONTAINER_IMAGE} ${CLUSTER_DOMAIN} ${ALLOW_PRIVILEGED} ${REGISTER_SCHEDULABLE} ${FAIL_SWAP_ON} ${ANONYMOUS_AUTH} ${IMAGE_PULL_PROGRESS_DEADLINE} ${HOSTNAME_OVERRIDE} ${EVICTION_HARD} ${AUTHENTICATION_TOKEN_WEBHOOK} ${CLIENT_CA_FILE} ${AUTHORIZATION_MODE} ${CLUSTER_DNS} ${NON_MASQUERADE_CIDR} ${NETWORK_PLUGIN} ${KUBECONFIG} ${V}
ExecStartPost=-/bin/bash /etc/kubernetes/deny-tcp-port-10250.sh
Restart=always
RestartSec=10
LimitNOFILE=65536

[Install]
WantedBy=multi-user.target

根據啓動文件發現環境文件在kubelet中

[root@VM_8_9_centos ~]# cat /etc/kubernetes/kubelet
CNI_BIN_DIR="--cni-bin-dir=/opt/cni/bin/"
KUBE_RESERVED="--kube-reserved=cpu=90m,memory=1830Mi"
MAX_PODS="--max-pods=253"
CLOUD_PROVIDER="--cloud-provider=qcloud"
CLOUD_CONFIG="--cloud-config=/etc/kubernetes/qcloud.conf"
POD_INFRA_CONTAINER_IMAGE="--pod-infra-container-image=ccr.ccs.tencentyun.com/library/pause:latest"
CLUSTER_DOMAIN="--cluster-domain=cluster.local"
ALLOW_PRIVILEGED="--allow-privileged=true"
REGISTER_SCHEDULABLE="--register-schedulable=true"
FAIL_SWAP_ON="--fail-swap-on=false"
ANONYMOUS_AUTH="--anonymous-auth=false"
IMAGE_PULL_PROGRESS_DEADLINE="--image-pull-progress-deadline=10m0s"
HOSTNAME_OVERRIDE="--hostname-override=192.168.8.9"
EVICTION_HARD="--eviction-hard=nodefs.available<10%,nodefs.inodesFree<5%,imagefs.available<15%,memory.available<100Mi"
AUTHENTICATION_TOKEN_WEBHOOK="--authentication-token-webhook=true"
CLIENT_CA_FILE="--client-ca-file=/etc/kubernetes/cluster-ca.crt"
AUTHORIZATION_MODE="--authorization-mode=Webhook"
CLUSTER_DNS="--cluster-dns=172.16.255.48"
NON_MASQUERADE_CIDR="--non-masquerade-cidr=0.0.0.0/0"
NETWORK_PLUGIN="--network-plugin=cni"
KUBECONFIG="--kubeconfig=/etc/kubernetes/kubelet-kubeconfig"
V="--v=2"

驅逐節點
驅逐會自動封鎖節點（將節點內pod驅逐到集羣中其餘節點上，不影響集羣使用，DaemonSet pod除外），驅逐後會檢查節點爲不可調度狀態

[root@VM_8_9_centos ~]# kubectl drain 192.168.8.9
node/192.168.8.9 cordoned
error: unable to drain node "192.168.8.9", aborting command...

There are pending nodes to be drained:
 192.168.8.9
error: DaemonSet-managed pods (use --ignore-daemonsets to ignore): rancher-logging-fluentd-tlqnq, rancher-logging-log-aggregator-m5g69, registry-proxy-xkmk5, exporter-node-cluster-monitoring-t5jj7, cattle-node-agent-j8wcn, ccs-log-collector-x57f8, ip-masq-agent-xt9q2, tke-bridge-agent-8rxgl, tke-cni-agent-lvv98; pods with local storage (use --delete-local-data to override): coupon-service-1, yuedi-passenger-api-1, coupon-service-1, rental-api-1, passenger-api-0, eureka-1, service-9777bd6ff-bsnj7, rental-service-0, passenger-api-0, swift-778d866966-c6nk7

[root@VM_8_9_centos ~]# kubectl get nodes
NAME        STATUS                     ROLES    AGE    VERSION
192.168.8.14   Ready                      <none>   89d    v1.12.4-tke.10
192.168.8.15   Ready                      <none>   181d   v1.12.4-tke.3
192.168.8.3    Ready                      <none>   181d   v1.12.4-tke.3
192.168.8.4    Ready                      <none>   181d   v1.12.4-tke.3
192.168.8.9    Ready,SchedulingDisabled   <none>   89d    v1.12.4-tke.10

[root@VM_8_9_centos ~]# ls /data/
docker  kubelet  lost+found

修改配置
修改kubelet配置文件第一行添加kubelet 目錄參數

[root@VM_8_9_centos ~]#  head -n 1 /etc/kubernetes/kubelet
ROOT_DIR="--root-dir=/data/kubelet"

修改啓動配置文件添加ROOT_DIR參數

[root@VM_8_9_centos ~]# vim /usr/lib/systemd/system/kubelet.service 
ExecStart=/usr/bin/kubelet ${ROOT_DIR} ${CNI_BIN_DIR} ${KUBE_RESERVED} ${MAX_PODS} ${CLOUD_PROVIDER} ${CLOUD_CONFIG} ${POD_INFRA_CONTAINER_IMAGE} ${CLUSTER_DOMAIN} ${ALLOW_PRIVILEGED} ${REGISTER_SCHEDULABLE} ${FAIL_SWAP_ON} ${ANONYMOUS_AUTH} ${IMAGE_PULL_PROGRESS_DEADLINE} ${HOSTNAME_OVERRIDE} ${EVICTION_HARD} ${AUTHENTICATION_TOKEN_WEBHOOK} ${CLIENT_CA_FILE} ${AUTHORIZATION_MODE} ${CLUSTER_DNS} ${NON_MASQUERADE_CIDR} ${NETWORK_PLUGIN} ${KUBECONFIG} ${V}

重啓服務

[root@VM_8_9_centos ~]# systemctl daemon-reload 
[root@VM_8_9_centos ~]# systemctl restart kubelet
[root@VM_8_9_centos ~]# systemctl status kubelet -l
● kubelet.service - kubelet
   Loaded: loaded (/usr/lib/systemd/system/kubelet.service; enabled; vendor preset: disabled)ls /
   Active: active (running) since Mon 2019-12-23 16:38:04 CST; 1min 2s ago
  Process: 24745 ExecStartPost=/bin/bash /etc/kubernetes/deny-tcp-port-10250.sh (code=exited, status=0/SUCCESS)
 Main PID: 24744 (kubelet)
    Tasks: 26
   Memory: 49.2M
   CGroup: /system.slice/kubelet.service
           └─24744 /usr/bin/kubelet --root-dir=/data/kubelet --cni-bin-dir=/opt/cni/bin/ --kube-reserved=cpu=90m,memory=1830Mi --max-pods=253 --cloud-provider=qcloud --cloud-config=/etc/kubernetes/qcloud.conf --pod-infra-container-image=ccr.ccs.tencentyun.com/library/pause:latest --cluster-domain=cluster.local --allow-privileged=true --register-schedulable=true --fail-swap-on=false --anonymous-auth=false --image-pull-progress-deadline=10m0s --hostname-override=192.168.8.9 --eviction-hard=nodefs.available<10%,nodefs.inodesFree<5%,imagefs.available<15%,memory.available<100Mi --authentication-token-webhook=true --client-ca-file=/etc/kubernetes/cluster-ca.crt --authorization-mode=Webhook --cluster-dns=172.16.255.48 --non-masquerade-cidr=0.0.0.0/0 --network-plugin=cni --kubeconfig=/etc/kubernetes/kubelet-kubeconfig  --v=2

取消封鎖

[root@VM_8_9_centos ~]# kubectl uncordon 192.168.8.9
node/192.168.8.9 uncordoned

驗證效果

查看節點信息正常

[root@VM_8_9_centos ~]# kubectl get nodes
NAME        STATUS   ROLES    AGE    VERSION
192.168.8.14   Ready    <none>   89d    v1.12.4-tke.10
192.168.8.15   Ready    <none>   181d   v1.12.4-tke.3
192.168.8.3    Ready    <none>   181d   v1.12.4-tke.3
192.168.8.4    Ready    <none>   181d   v1.12.4-tke.3
192.168.8.9    Ready    <none>   89d    v1.12.4-tke.10

新建kubelet目錄在data下

[root@VM_8_9_centos ~]# ls /data/kubelet/
cpu_manager_state  plugin-containers  plugins  pods
[root@VM_8_9_centos ~]# du -sh /data/kubelet/
480K    /data/kubelet/