kubernetes集群升级

k8s集群master节点升级和node节点升级

soliso

547人浏览 · 2022-11-13 17:26:02

soliso · 2022-11-13 17:26:02 发布

一、下载

1、搜索kubernetes，而后点击kubernetes

1、先查看目前集群版本
root@k8s-deploy:~# kubectl get node
NAME           STATUS                     ROLES    AGE     VERSION
172.31.7.101   Ready,SchedulingDisabled   master   5h22m   v1.24.2
172.31.7.102   Ready,SchedulingDisabled   master   5h22m   v1.24.2
172.31.7.103   Ready,SchedulingDisabled   master   3h8m    v1.24.2
172.31.7.111   Ready                      node     5h20m   v1.24.2
172.31.7.112   Ready                      node     5h20m   v1.24.2
172.31.7.113   Ready                      node     3h22m   v1.24.2
root@k8s-deploy:~#

注意事项：集群升级最好是跨小版本，不要跨大版本升级，比如1.24.2可以升级到1.24.X但是最好不要直接升级到1.25版本，因为不同版本的server文件未必是百分百兼容的升级方式主要是通过替换二进制实现的，把api server的二进制，kubelet、kubeporxy等二进制替换掉实现升级

一、下载

进入github官网，下载升级需要用的新版本包

1、搜索kubernetes，而后点击kubernetes

2、基于上一步进入的页面右下框，点击版本

3、找到需要升级的版本1.24.3

4、点击下载

5、下载升级需要的四个包文件

下载的链接和截图

https://dl.k8s.io/v1.24.3/kubernetes.tar.gz

https://dl.k8s.io/v1.24.3/kubernetes-client-linux-amd64.tar.gz

下载链接和截图

https://dl.k8s.io/v1.24.3/kubernetes-server-linux-amd64.tar.gz

下载链接和截图

https://dl.k8s.io/v1.24.3/kubernetes-node-linux-amd64.tar.gz

二、上传

把下载好的四个包文件上传至部署服务器

1、上传到部署服务器
root@k8s-deploy:~# ls
ezdown      kubernetes-client-linux-amd64.tar.gz  kubernetes-server-linux-amd64.tar.gz  ls
kubernetes-node-linux-amd64.tar.gz    kubernetes.tar.gz                     snap

2、解压
root@k8s-deploy:~# tar xf kubernetes-client-linux-amd64.tar.gz
root@k8s-deploy:~# tar xf kubernetes-node-linux-amd64.tar.gz
root@k8s-deploy:~# tar xf kubernetes-server-linux-amd64.tar.gz
root@k8s-deploy:~# tar xf kubernetes.tar.gz
解压后有个kubernetes目录
root@k8s-deploy:~# ls
ezdown      kubernetes-client-linux-amd64.tar.gz  kubernetes-server-linux-amd64.tar.gz  ls
kubernetes  kubernetes-node-linux-amd64.tar.gz    kubernetes.tar.gz                     snap

root@k8s-deploy:~# cd kubernetes/
在/kubernetes/cluster/addons/目录中有很多插件
root@k8s-deploy:~/kubernetes# ll cluster/addons/
total 80
drwxr-xr-x 18 root root 4096 Jul 13 14:48 ./
drwxr-xr-x  9 root root 4096 Jul 13 14:48 ../
drwxr-xr-x  2 root root 4096 Jul 13 14:48 addon-manager/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 calico-policy-controller/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 cluster-loadbalancing/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 device-plugins/
drwxr-xr-x  5 root root 4096 Jul 13 14:48 dns/
drwxr-xr-x  2 root root 4096 Jul 13 14:48 dns-horizontal-autoscaler/
drwxr-xr-x  4 root root 4096 Jul 13 14:48 fluentd-gcp/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 ip-masq-agent/
drwxr-xr-x  2 root root 4096 Jul 13 14:48 kube-proxy/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 metadata-agent/
drwxr-xr-x  3 root root 4096 Jul 13 14:48 metadata-proxy/
drwxr-xr-x  2 root root 4096 Jul 13 14:48 metrics-server/
drwxr-xr-x  5 root root 4096 Jul 13 14:48 node-problem-detector/
-rw-r--r--  1 root root  104 Jul 13 14:48 OWNERS
drwxr-xr-x  8 root root 4096 Jul 13 14:48 rbac/
-rw-r--r--  1 root root 1655 Jul 13 14:48 README.md
drwxr-xr-x  8 root root 4096 Jul 13 14:48 storage-class/
drwxr-xr-x  4 root root 4096 Jul 13 14:48 volumesnapshots/

比如dns目录里面有官方提供的yaml文件，我们用coredns的话可以再到下一级目录看下
root@k8s-deploy:~/kubernetes# ll cluster/addons/dns
total 24
drwxr-xr-x  5 root root 4096 Jul 13 14:48 ./
drwxr-xr-x 18 root root 4096 Jul 13 14:48 ../
drwxr-xr-x  2 root root 4096 Jul 13 14:48 coredns/
drwxr-xr-x  2 root root 4096 Jul 13 14:48 kube-dns/
drwxr-xr-x  2 root root 4096 Jul 13 14:48 nodelocaldns/
-rw-r--r--  1 root root  141 Jul 13 14:48 OWNERS

只需要把coredns目录里的yaml文件拷贝后修改就行
root@k8s-deploy:~/kubernetes# ll cluster/addons/dns/coredns/
total 44
drwxr-xr-x 2 root root 4096 Jul 13 14:48 ./
drwxr-xr-x 5 root root 4096 Jul 13 14:48 ../
-rw-r--r-- 1 root root 5060 Jul 13 14:48 coredns.yaml.base
-rw-r--r-- 1 root root 5110 Jul 13 14:48 coredns.yaml.in
-rw-r--r-- 1 root root 5112 Jul 13 14:48 coredns.yaml.sed
-rw-r--r-- 1 root root 1075 Jul 13 14:48 Makefile
-rw-r--r-- 1 root root  344 Jul 13 14:48 transforms2salt.sed
-rw-r--r-- 1 root root  287 Jul 13 14:48 transforms2sed.sed
这里暂时不做修改，后面会用到


找以下二进制文件kube-apiserver、kube-controller-manager、kube-scheduler、kube-proxy、kubectl、kubelet主要这几个二进制这几个二进制都是官方已经编译好的
root@k8s-deploy:~/kubernetes# ll server/bin/
total 1089112
drwxr-xr-x 2 root root      4096 Jul 13 14:43 ./
drwxr-xr-x 3 root root      4096 Jul 13 14:48 ../
-rwxr-xr-x 1 root root  55402496 Jul 13 14:43 apiextensions-apiserver*
-rwxr-xr-x 1 root root  44376064 Jul 13 14:43 kubeadm*
-rwxr-xr-x 1 root root  49496064 Jul 13 14:43 kube-aggregator*
-rwxr-xr-x 1 root root 125865984 Jul 13 14:43 kube-apiserver*
-rw-r--r-- 1 root root         8 Jul 13 14:41 kube-apiserver.docker_tag
-rw------- 1 root root 131066880 Jul 13 14:41 kube-apiserver.tar
-rwxr-xr-x 1 root root 115515392 Jul 13 14:43 kube-controller-manager*
-rw-r--r-- 1 root root         8 Jul 13 14:41 kube-controller-manager.docker_tag
-rw------- 1 root root 120716288 Jul 13 14:41 kube-controller-manager.tar
-rwxr-xr-x 1 root root  45711360 Jul 13 14:43 kubectl*
-rwxr-xr-x 1 root root  55036584 Jul 13 14:43 kubectl-convert*
-rwxr-xr-x 1 root root 116013432 Jul 13 14:43 kubelet*
-rwxr-xr-x 1 root root   1486848 Jul 13 14:43 kube-log-runner*
-rwxr-xr-x 1 root root  41762816 Jul 13 14:43 kube-proxy*
-rw-r--r-- 1 root root         8 Jul 13 14:41 kube-proxy.docker_tag
-rw------- 1 root root 111863808 Jul 13 14:41 kube-proxy.tar
-rwxr-xr-x 1 root root  47144960 Jul 13 14:43 kube-scheduler*
-rw-r--r-- 1 root root         8 Jul 13 14:41 kube-scheduler.docker_tag
-rw------- 1 root root  52345856 Jul 13 14:41 kube-scheduler.tar
-rwxr-xr-x 1 root root   1413120 Jul 13 14:43 mounter*

检查下版本，是否我们需要的版本
root@k8s-deploy:~/kubernetes# ./server/bin/kube-apiserver --version
Kubernetes v1.24.3

三、升级

1、升级master1

升级需要把所有node节点上的负载均衡器中的api server拿掉,这样就不会有请求再转给那台机器了
下面先升级master1  172.31.7.101


1、node1修改配置文件
root@node1:~# vim /etc/kube-lb/conf/kube-lb.conf
把升级的节点注释掉
upstream backend {
        server 172.31.7.103:6443    max_fails=2 fail_timeout=3s;
#        server 172.31.7.101:6443    max_fails=2 fail_timeout=3s;
        server 172.31.7.102:6443    max_fails=2 fail_timeout=3s;
    }

root@node1:~# systemctl reload kube-lb.service

2、node2拿掉api server
root@node2:~# sed -i 's/server 172.31.7.101/#server 172.31.7.101/' /etc/kube-lb/conf/kube-lb.conf
root@node2:~# systemctl reload kube-lb.service

3、node3拿掉api server
root@node3:~# sed -i 's/server 172.31.7.101/#server 172.31.7.101/' /etc/kube-lb/conf/kube-lb.conf
root@node2:~# systemctl reload kube-lb.service
以上做完后就不会有请求再转给master1了，就相当于这台api server服务器已经下线了

然后把master1的一些管理端服务关掉，不然现在进程已经正在占用二进制了
关停下面五个服务
kube-apiserver.service 
kube-controller-manager.service 
kubelet.service 
kube-proxy.service
kube-scheduler.service

4、关闭服务
root@master1:~# systemctl stop kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service
root@master1:~#


这几个二进制文件都是在/usr/local/bin/目录下
root@master1:~# ll /usr/local/bin/
total 826520
drwxr-xr-x  2 root root      4096 Nov 12 12:02 ./
drwxr-xr-x 10 root root      4096 Aug 24  2021 ../
-rwxr-xr-x  1 root root   4159518 Nov 12 12:02 bandwidth*
-rwxr-xr-x  1 root root   4221977 Nov 11 07:46 bridge*
-rwxr-xr-x  1 root root  41472000 Nov 12 12:02 calico*
-rwxr-xr-x  1 root root  44978176 Nov 11 08:51 calicoctl*
-rwxr-xr-x  1 root root  41472000 Nov 12 12:02 calico-ipam*
-rwxr-xr-x  1 root root  59584224 Nov 11 03:51 containerd*
-rwxr-xr-x  1 root root   7389184 Nov 11 03:51 containerd-shim*
-rwxr-xr-x  1 root root   9555968 Nov 11 03:51 containerd-shim-runc-v1*
-rwxr-xr-x  1 root root   9580544 Nov 11 03:51 containerd-shim-runc-v2*
-rwxr-xr-x  1 root root  33228961 Nov 11 03:52 crictl*
-rwxr-xr-x  1 root root  30256896 Nov 11 03:52 ctr*
-rwxr-xr-x  1 root root   3069556 Nov 12 12:02 flannel*
-rwxr-xr-x  1 root root   3614480 Nov 12 12:02 host-local*
-rwxr-xr-x  1 root root  41472000 Nov 12 12:02 install*
-rwxr-xr-x  1 root root 125865984 Nov 11 07:45 kube-apiserver*
-rwxr-xr-x  1 root root 115507200 Nov 11 07:45 kube-controller-manager*
-rwxr-xr-x  1 root root  45711360 Nov 11 07:45 kubectl*
-rwxr-xr-x  1 root root 116353400 Nov 11 07:46 kubelet*
-rwxr-xr-x  1 root root  41762816 Nov 11 07:46 kube-proxy*
-rwxr-xr-x  1 root root  47144960 Nov 11 07:45 kube-scheduler*
-rwxr-xr-x  1 root root   3209463 Nov 12 12:02 loopback*
-rwxr-xr-x  1 root root   3939867 Nov 12 12:02 portmap*
-rwxr-xr-x  1 root root   9419136 Nov 11 03:52 runc*
-rwxr-xr-x  1 root root   3356587 Nov 12 12:02 tuning*
目前都是v1.24.2版本
root@master1:~# /usr/local/bin/kube-apiserver --version
Kubernetes v1.24.2


把新的二进制文件拷贝到/usr/local/bin/即可完成升级
5、拷贝二进制，进行升级
root@k8s-deploy:~/kubernetes/server/bin# scp kube-apiserver kube-controller-manager kube-scheduler kube-proxy kubelet  kubectl 172.31.7.101:/usr/local/bin/
kube-apiserver                                                             100%  120MB  74.6MB/s   00:01
kube-controller-manager                                                    100%  110MB  80.7MB/s   00:01
kube-scheduler                                                             100%   45MB  54.7MB/s   00:00
kube-proxy                                                                 100%   40MB  73.5MB/s   00:00
kubelet                                                                    100%  111MB  60.1MB/s   00:01
kubectl                                                                    100%   44MB  68.0MB/s   00:00

6、此时再到master1上验证下版本，是否已经升级了
root@master1:~# /usr/local/bin/kube-apiserver --version
Kubernetes v1.24.3

7、启动服务
root@master1:~# systemctl start  kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service
root@master1:~#

8、可以再看下集群版本信息了，172.31.7.101已经是v1.24.3版本了
root@k8s-deploy:~/kubernetes/server/bin# kubectl get node
NAME           STATUS                        ROLES    AGE     VERSION
172.31.7.101   Ready,SchedulingDisabled      master   6h36m   v1.24.3
172.31.7.102   Ready,SchedulingDisabled      master   6h36m   v1.24.2
172.31.7.103   Ready,SchedulingDisabled      master   4h21m   v1.24.2
172.31.7.111   Ready                         node     6h33m   v1.24.2
172.31.7.112   Ready                         node     6h33m   v1.24.2
172.31.7.113   Ready                         node     4h36m   v1.24.2


这个版本是由kubelet决定的，只要把kubelet更新了，哪怕没有更新kube-proxy也会显示1.24.3版本

2、升级master2,master3

1、把master2.master3拿掉，此时node节点还可以通过master1正常工作
root@node1:~# vim /etc/kube-lb/conf/kube-lb.conf
#        server 172.31.7.103:6443    max_fails=2 fail_timeout=3s;
        server 172.31.7.101:6443    max_fails=2 fail_timeout=3s;
#        server 172.31.7.102:6443    max_fails=2 fail_timeout=3s;

root@node1:~# systemctl reload kube-lb.service


2、其它两个node节点也依次修改，重启服务

把master1,master2五个服务都停掉
root@master2:~# systemctl stop  kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service
root@master3:~# systemctl stop  kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service

3、把部署节点上的五个新版本二进制文件拷贝到master1,master2
root@k8s-deploy:~/kubernetes/server/bin# scp kube-apiserver kube-controller-manager kube-scheduler kube-proxy                        kubelet  kubectl 172.31.7.102:/usr/local/bin/
root@k8s-deploy:~/kubernetes/server/bin# scp kube-apiserver kube-controller-manager kube-scheduler kube-proxy                        kubelet  kubectl 172.31.7.102:/usr/local/bin/

4、文件拷贝好后，把master1,master2上的服务都启动
root@master2:~# systemctl start  kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service

root@master3:~# systemctl start  kube-apiserver.service kube-controller-manager.service kubelet.service kube-proxy.service kube-scheduler.service

5、查看集群版本信息三个master节点都完成了升级
root@k8s-deploy:~/kubernetes/server/bin# kubectl get node
NAME           STATUS                        ROLES    AGE     VERSION
172.31.7.101   Ready,SchedulingDisabled      master   6h56m   v1.24.3
172.31.7.102   Ready,SchedulingDisabled      master   6h56m   v1.24.3
172.31.7.103   Ready,SchedulingDisabled      master   4h41m   v1.24.3
172.31.7.111   Ready                         node     6h53m   v1.24.2
172.31.7.112   Ready                         node     6h53m   v1.24.2
172.31.7.113   Ready                         node     4h56m   v1.24.2

3、升级node节点

node升级和master还不太一样，因为master上面没有运行业务容器，但是node上是运行业务容器的
所以要先把node节点驱逐出k8s环境，驱逐后node就没有业务容器了，就可以做升级操作了

需要驱逐node
每个node节点上都有daemonset容器，没法驱逐比如calico，且每个node节点上都有这个pod，因此无法驱逐到其它节点上运行，但是这种pod也不需要重建，因为每个节点都有了，而它无法驱逐，所以可以把这pod忽略掉
root@k8s-deploy:~/kubernetes/server/bin# kubectl get pod -o wide -n kube-system
NAME                                       READY   STATUS    RESTARTS        AGE     IP             NODE           NOMINATED NODE   READINESS GATES
calico-kube-controllers-69f77c86d8-2bzzt   1/1     Running   1 (103m ago)    7h1m    172.31.7.111   172.31.7.111   <none>           <none>
calico-node-9cg2k                          1/1     Running   1 (4h59m ago)   4h59m   172.31.7.111   172.31.7.111   <none>           <none>
calico-node-dctqp                          1/1     Running   1 (102m ago)    5h1m    172.31.7.102   172.31.7.102   <none>           <none>
calico-node-l7zf8                          1/1     Running   0               5h4m    172.31.7.113   172.31.7.113   <none>           <none>
calico-node-qw6c6                          1/1     Running   1 (10m ago)     4h50m   172.31.7.103   172.31.7.103   <none>           <none>
calico-node-rcvkr                          1/1     Running   2 (102m ago)    4h59m   172.31.7.101   172.31.7.101   <none>           <none>
calico-node-vgvzm                          1/1     Running   0               5h4m    172.31.7.112   172.31.7.112   <none>           <none>


1、驱逐node
root@k8s-deploy:~# kubectl drain 172.31.7.111 --ignore-daemonsets --force
root@k8s-deploy:~# kubectl get node
NAME           STATUS                     ROLES    AGE   VERSION
172.31.7.101   Ready,SchedulingDisabled   master   25h   v1.24.3
172.31.7.102   Ready,SchedulingDisabled   master   25h   v1.24.3
172.31.7.103   Ready,SchedulingDisabled   master   23h   v1.24.3
172.31.7.111   Ready,SchedulingDisabled   node     25h   v1.24.2
172.31.7.112   Ready                      node     25h   v1.24.2
172.31.7.113   Ready                      node     23h   v1.24.2
这时172.31.7.111也被打上SchedulingDisabled的标签了，业务容器都迁移了，不会再向这个主机调度新的pod


2、node节点停服务，升级
root@node1:~# systemctl stop kubelet.service  kube-proxy.service

3、拷贝新版本的二进制文件到node节点上
root@k8s-deploy:~/kubernetes/server/bin# scp kubelet kube-proxy kubectl 172.31.7.111:/usr/local/bin/

4、启动服务
root@node1:~# systemctl start  kubelet.service  kube-proxy.service

5、查看node1版本
root@node1:~# kubectl get node
NAME           STATUS                     ROLES    AGE   VERSION
172.31.7.101   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.102   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.103   Ready,SchedulingDisabled   master   23h   v1.24.3
172.31.7.111   Ready,SchedulingDisabled   node     25h   v1.24.3
172.31.7.112   Ready                      node     25h   v1.24.2
172.31.7.113   Ready                      node     24h   v1.24.2

6、恢复node1的调度状态，让它恢复正常
root@k8s-deploy:~/kubernetes/server/bin# kubectl uncordon 172.31.7.111
node/172.31.7.111 uncordoned

7、查看集群节点调度状态
root@k8s-deploy:~/kubernetes/server/bin# kubectl get node
NAME           STATUS                     ROLES    AGE   VERSION
172.31.7.101   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.102   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.103   Ready,SchedulingDisabled   master   23h   v1.24.3
172.31.7.111   Ready                      node     26h   v1.24.3
172.31.7.112   Ready                      node     26h   v1.24.2
172.31.7.113   Ready                      node     24h   v1.24.2

升级其它两个node
以上以生产环境有业务情况下比较严谨的先驱逐再升级，下面两个node我就不驱逐直接升级了首先停服务，然后把新的二进制拷贝过去，最后再启动服务
1、node2节点升级
root@node2:~# systemctl stop kubelet.service  kube-proxy.service
root@k8s-deploy:~/kubernetes/server/bin# scp kubelet kube-proxy kubectl 172.31.7.112:/usr/local/bin/
root@node2:~# systemctl start kubelet.service  kube-proxy.service

2、node3节点升级
root@node3:~# systemctl stop kubelet.service  kube-proxy.service
root@k8s-deploy:~/kubernetes/server/bin# scp kubelet kube-proxy kubectl 172.31.7.113:/usr/local/bin/
root@node3:~# systemctl start kubelet.service  kube-proxy.service

查看集群版本号
root@node3:~# kubectl get node
NAME           STATUS                     ROLES    AGE   VERSION
172.31.7.101   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.102   Ready,SchedulingDisabled   master   26h   v1.24.3
172.31.7.103   Ready,SchedulingDisabled   master   23h   v1.24.3
172.31.7.111   Ready                      node     26h   v1.24.3
172.31.7.112   Ready                      node     26h   v1.24.3
172.31.7.113   Ready                      node     24h   v1.24.3

集群升级完成，再把/etc/kube-lb/conf/kube-lb.conf文件中api server注释都解开
root@node1:~# vim /etc/kube-lb/conf/kube-lb.conf
root@node1:~# systemctl  reload kube-lb.service

root@node2:~# vim /etc/kube-lb/conf/kube-lb.conf
root@node2:~# systemctl  reload kube-lb.service

root@node3:~# vim /etc/kube-lb/conf/kube-lb.conf
root@node3:~# systemctl  reload kube-lb.service

最后把这些二进制拷贝到/etc/kubease/bin/目录下，这样即使后面新加的节点也会使用这些新的二进制了，也都是新版本了

root@k8s-deploy:~/kubernetes/server/bin# \cp kube-apiserver kube-controller-manager kube-scheduler kube-proxy kubelet kubectl /etc/kubeasz/bin/

至此就完成了master,node的升级，希望对你有所帮助，谢谢~

K8S/Kubernetes

K8S/Kubernetes社区为您提供最前沿的新闻资讯和知识内容

更多推荐

【深度】阿里巴巴万级规模 K8s 集群全局高可用体系之美

作者 | 韩堂、柘远、沉醉来源 | 阿里巴巴云原生公众号前言台湾作家林清玄在接受记者采访的时候，如此评价自己 30 多年写作生涯：“第一个十年我才华横溢，‘贼光闪现’，令周边黯然失色；第二个十年，我终于‘宝光现形’，不再去抢风头，反而与身边的美丽相得益彰；进入第三个十年，繁华落尽见真醇，我进入了‘醇光初现’的阶段，真正体味到了境界之美”。长夜有穷，真水无香。领略过了 K8s“身在江

K8S/Kubernetes

如何基于 K8s 构建下一代 DevOps 平台？

作者 | 孙健波（天元）导读：当前云原生 DevOps 体系现状如何？面临哪些挑战？如何通过 OAM 解决云原生 DevOps 场景下的诸多问题？云原生开发应用模型 OAM(Open Application Model) 社区核心成员孙健波将为大家一一解答，并分享如何基于 OAM 和 Kubernetes 打造无限能力的下一代 DevOps 平台。什么是 DevOps？为什么基于 Kub