1. 环境准备
1.1. 部署机器
以下机器为虚拟机
机器IP
主机名
角色
系统版本
备注
172.16.94.140
kube-master-0
k8s master
Centos 4.17.14
内存：3G
172.16.94.141
kube-node-41
k8s node
Centos 4.17.14
内存：3G
172.16.94.142
kube-node-42
k8s node
Centos 4.17.14
内存：3G
172.16.94.135
部署管理机
-
1.2. 配置管理机
管理机主要用来部署k8s集群，需要安装以下版本的软件，具体可参考：
1 2 3 4 5 6 ansible>=2.4.0 jinja2>=2.9.6 netaddr pbr>=1.6 ansible-modules-hashivault>=3.9.4 hvac
1、安装及配置ansible
2、安装python-netaddr
1 2 3 4 5 yum -y install epel-release yum -y install python-pip pip install netaddr
3、升级Jinja
1 2 pip install --upgrade jinja2
1.3. 配置部署机器
部署机器即用来运行k8s集群的机器，包括Master和Node。
1、确认系统版本
本文采用centos7的系统，建议将系统内核升级到4.x.x以上。
2、关闭防火墙
1 2 3 systemctl stop firewalld systemctl disable firewalld iptables -F
3、关闭swap
Kubespary v2.5.0的版本需要关闭swap，具体参考
1 2 3 4 5 - name: Stop if swap enabled assert: that: ansible_swaptotal_mb == 0 when: kubelet_fail_swap_on|default(true) ignore_errors: "{{ ignore_assert_errors }} "
V2.6.0 版本去除了swap的检查，具体参考：
执行关闭swap命令swapoff -a。
1 2 3 4 5 6 7 8 [root@master ~] [root@master ~] [root@master ~] total used free shared buff/cache available Mem: 976 366 135 6 474 393 Swap: 0 0 0
4、确认部署机器内存
由于本文采用虚拟机部署，内存可能存在不足的问题，因此将虚拟机内存调整为3G或以上；如果是物理机一般不会有内存不足的问题。具体参考：
1 2 3 4 5 6 7 8 9 10 11 - name: Stop if memory is too small for masters assert: that: ansible_memtotal_mb >= 1500 ignore_errors: "{{ ignore_assert_errors }} " when: inventory_hostname in groups['kube-master'] - name: Stop if memory is too small for nodes assert: that: ansible_memtotal_mb >= 1024 ignore_errors: "{{ ignore_assert_errors }} " when: inventory_hostname in groups['kube-node']
1.4. 涉及镜像
Docker版本为17.03.2-ce。
1、Master节点
2、Node节点
3、说明
镜像被墙并且全部镜像下载需要较多时间，建议提前下载到部署机器上。
hyperkube镜像主要用来运行k8s核心组件（例如kube-apiserver等）。
此处使用的网络组件为calico。
2. 部署集群
2.1. 下载kubespary的源码
1 git clone https://github.com/kubernetes-incubator/kubespray.git
2.2. 编辑配置文件
2.2.1. hosts.ini
hosts.ini主要为部署节点机器信息的文件，路径为：kubespray/inventory/sample/hosts.ini。
1 2 3 4 cd kubespraycp -rfp inventory/sample inventory/k8s vi inventory/k8s/hosts.ini
例如：
hosts.ini文件可以填写部署机器的登录密码，也可以不填密码而设置ssh的免密登录。
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 kube-master-0 ansible_ssh_host=172.16.94.140 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.140 mask=/24 kube-node-41 ansible_ssh_host=172.16.94.141 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.141 mask=/24 kube-node-42 ansible_ssh_host=172.16.94.142 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.142 mask=/24 [kube-master] kube-master-0 [etcd] kube-master-0 [kube-node] kube-node-41 kube-node-42 [k8s-cluster:children] kube-node kube-master [calico-rr]
2.2.2. k8s-cluster.yml
k8s-cluster.yml主要为k8s集群的配置文件，路径为：kubespray/inventory/k8s/group_vars/k8s-cluster.yml。该文件可以修改安装的k8s集群的版本，参数为：kube_version: v1.9.5。具体可参考：
2.3. 执行部署操作
涉及文件为cluster.yml。
1 2 3 4 cd kubesprayansible-playbook -i inventory/k8s/hosts.ini cluster.yml -b -vvv
-vvv 参数表示输出运行日志
如果需要重置可以执行以下命令：
涉及文件为reset.yml。
1 ansible-playbook -i inventory/k8s/hosts.ini reset.yml -b -vvv
3. 确认部署结果
3.1. ansible的部署结果
ansible命令执行完，出现以下日志，则说明部署成功，否则根据报错内容进行修改。
1 2 3 4 5 PLAY RECAP ***************************************************************************** kube-master-0 : ok=309 changed=30 unreachable=0 failed=0 kube-node-41 : ok=203 changed=8 unreachable=0 failed=0 kube-node-42 : ok=203 changed=8 unreachable=0 failed=0 localhost : ok=2 changed=0 unreachable=0 failed=0
以下为部分部署执行日志：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 kubernetes/preinstall : Update package management cache (YUM) --------------------23.96s /root/gopath/src/kubespray/roles/kubernetes/preinstall/tasks/main.yml:121 kubernetes/master : Master | wait for the apiserver to be running ----------------23.44s /root/gopath/src/kubespray/roles/kubernetes/master/handlers/main.yml:79 kubernetes/preinstall : Install packages requirements ----------------------------20.20s /root/gopath/src/kubespray/roles/kubernetes/preinstall/tasks/main.yml:203 kubernetes/secrets : Check certs | check if a cert already exists on node --------13.94s /root/gopath/src/kubespray/roles/kubernetes/secrets/tasks/check-certs.yml:17 gather facts from all instances --------------------------------------------------9.98s /root/gopath/src/kubespray/cluster.yml:25 kubernetes/node : install | Compare host kubelet with hyperkube container --------9.66s /root/gopath/src/kubespray/roles/kubernetes/node/tasks/install_host.yml:2 kubernetes-apps/ansible : Kubernetes Apps | Start Resources -----------------------9.27s /root/gopath/src/kubespray/roles/kubernetes-apps/ansible/tasks/main.yml:37 kubernetes-apps/ansible : Kubernetes Apps | Lay Down KubeDNS Template ------------8.47s /root/gopath/src/kubespray/roles/kubernetes-apps/ansible/tasks/kubedns.yml:3 download : Sync container ---------------------------------------------------------8.23s /root/gopath/src/kubespray/roles/download/tasks/main.yml:15 kubernetes-apps/network_plugin/calico : Start Calico resources --------------------7.82s /root/gopath/src/kubespray/roles/kubernetes-apps/network_plugin/calico/tasks/main.yml:2 download : Download items ---------------------------------------------------------7.67s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 download : Download items ---------------------------------------------------------7.48s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 download : Sync container ---------------------------------------------------------7.35s /root/gopath/src/kubespray/roles/download/tasks/main.yml:15 download : Download items ---------------------------------------------------------7.16s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 network_plugin/calico : Calico | Copy cni plugins from calico/cni container -------7.10s /root/gopath/src/kubespray/roles/network_plugin/calico/tasks/main.yml:62 download : Download items ---------------------------------------------------------7.04s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 download : Download items ---------------------------------------------------------7.01s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 download : Sync container ---------------------------------------------------------7.00s /root/gopath/src/kubespray/roles/download/tasks/main.yml:15 download : Download items ---------------------------------------------------------6.98s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6 download : Download items ---------------------------------------------------------6.79s /root/gopath/src/kubespray/roles/download/tasks/main.yml:6
3.2. k8s集群运行结果
1、k8s组件信息
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE ds/calico-node 3 3 3 3 3 <none> 2h NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE deploy/kube-dns 2 2 2 2 2h deploy/kubedns-autoscaler 1 1 1 1 2h deploy/kubernetes-dashboard 1 1 1 1 2h NAME DESIRED CURRENT READY AGE rs/kube-dns-79d99cdcd5 2 2 2 2h rs/kubedns-autoscaler-5564b5585f 1 1 1 2h rs/kubernetes-dashboard-69cb58d748 1 1 1 2h NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE ds/calico-node 3 3 3 3 3 <none> 2h NAME DESIRED CURRENT UP-TO-DATE AVAILABLE AGE deploy/kube-dns 2 2 2 2 2h deploy/kubedns-autoscaler 1 1 1 1 2h deploy/kubernetes-dashboard 1 1 1 1 2h NAME DESIRED CURRENT READY AGE rs/kube-dns-79d99cdcd5 2 2 2 2h rs/kubedns-autoscaler-5564b5585f 1 1 1 2h rs/kubernetes-dashboard-69cb58d748 1 1 1 2h NAME READY STATUS RESTARTS AGE po/calico-node-22vsg 1/1 Running 0 2h po/calico-node-t7zgw 1/1 Running 0 2h po/calico-node-zqnx8 1/1 Running 0 2h po/kube-apiserver-kube-master-0 1/1 Running 0 22h po/kube-controller-manager-kube-master-0 1/1 Running 0 2h po/kube-dns-79d99cdcd5-f2t6t 3/3 Running 0 2h po/kube-dns-79d99cdcd5-gw944 3/3 Running 0 2h po/kube-proxy-kube-master-0 1/1 Running 2 22h po/kube-proxy-kube-node-41 1/1 Running 3 22h po/kube-proxy-kube-node-42 1/1 Running 3 22h po/kube-scheduler-kube-master-0 1/1 Running 0 2h po/kubedns-autoscaler-5564b5585f-lt9bb 1/1 Running 0 2h po/kubernetes-dashboard-69cb58d748-wmb9x 1/1 Running 0 2h po/nginx-proxy-kube-node-41 1/1 Running 3 22h po/nginx-proxy-kube-node-42 1/1 Running 3 22h NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE svc/kube-dns ClusterIP 10.233.0.3 <none> 53/UDP,53/TCP 2h svc/kubernetes-dashboard ClusterIP 10.233.27.24 <none> 443/TCP 2h
2、k8s节点信息
1 2 3 4 5 NAME STATUS ROLES AGE VERSION kube-master-0 Ready master 22h v1.9.5 kube-node-41 Ready node 22h v1.9.5 kube-node-42 Ready node 22h v1.9.5
3、组件健康信息
1 2 3 4 5 NAME STATUS MESSAGE ERROR scheduler Healthy ok controller-manager Healthy ok etcd-0 Healthy {"health" : "true" }
4. k8s集群扩容节点
4.1. 修改hosts.ini文件
如果需要扩容Node节点，则修改hosts.ini文件，增加新增的机器信息。例如，要增加节点机器kube-node-43（IP为172.16.94.143），修改后的文件内容如下：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 kube-master-0 ansible_ssh_host=172.16.94.140 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.140 mask=/24 kube-node-41 ansible_ssh_host=172.16.94.141 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.141 mask=/24 kube-node-42 ansible_ssh_host=172.16.94.142 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.142 mask=/24 kube-node-43 ansible_ssh_host=172.16.94.143 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.143 mask=/24 [kube-master] kube-master-0 [etcd] kube-master-0 [kube-node] kube-node-41 kube-node-42 kube-node-43 [k8s-cluster:children] kube-node kube-master [calico-rr]
4.2. 执行扩容命令
涉及文件为scale.yml。
1 2 3 4 cd kubesprayansible-playbook -i inventory/k8s/hosts.ini scale.yml -b -vvv
4.3. 检查扩容结果
1、ansible的执行结果
1 2 3 4 5 PLAY RECAP *************************************** kube-node-41 : ok=228 changed=11 unreachable=0 failed=0 kube-node-42 : ok=197 changed=6 unreachable=0 failed=0 kube-node-43 : ok=227 changed=69 unreachable=0 failed=0 localhost : ok=2 changed=0 unreachable=0 failed=0
2、k8s的节点信息
1 2 3 4 5 6 NAME STATUS ROLES AGE VERSION kube-master-0 Ready master 1d v1.9.5 kube-node-41 Ready node 1d v1.9.5 kube-node-42 Ready node 1d v1.9.5 kube-node-43 Ready node 1m v1.9.5
可以看到新增的kube-node-43节点已经扩容完成。
3、k8s组件信息
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 NAME READY STATUS RESTARTS AGE IP NODE calico-node-22vsg 1/1 Running 0 10h 172.16.94.140 kube-master-0 calico-node-8fz9x 1/1 Running 2 27m 172.16.94.143 kube-node-43 calico-node-t7zgw 1/1 Running 0 10h 172.16.94.142 kube-node-42 calico-node-zqnx8 1/1 Running 0 10h 172.16.94.141 kube-node-41 kube-apiserver-kube-master-0 1/1 Running 0 1d 172.16.94.140 kube-master-0 kube-controller-manager-kube-master-0 1/1 Running 0 10h 172.16.94.140 kube-master-0 kube-dns-79d99cdcd5-f2t6t 3/3 Running 0 10h 10.233.100.194 kube-node-41 kube-dns-79d99cdcd5-gw944 3/3 Running 0 10h 10.233.107.1 kube-node-42 kube-proxy-kube-master-0 1/1 Running 2 1d 172.16.94.140 kube-master-0 kube-proxy-kube-node-41 1/1 Running 3 1d 172.16.94.141 kube-node-41 kube-proxy-kube-node-42 1/1 Running 3 1d 172.16.94.142 kube-node-42 kube-proxy-kube-node-43 1/1 Running 0 26m 172.16.94.143 kube-node-43 kube-scheduler-kube-master-0 1/1 Running 0 10h 172.16.94.140 kube-master-0 kubedns-autoscaler-5564b5585f-lt9bb 1/1 Running 0 10h 10.233.100.193 kube-node-41 kubernetes-dashboard-69cb58d748-wmb9x 1/1 Running 0 10h 10.233.107.2 kube-node-42 nginx-proxy-kube-node-41 1/1 Running 3 1d 172.16.94.141 kube-node-41 nginx-proxy-kube-node-42 1/1 Running 3 1d 172.16.94.142 kube-node-42 nginx-proxy-kube-node-43 1/1 Running 0 26m 172.16.94.143 kube-node-43
5. 部署高可用集群
将hosts.ini文件中的master和etcd的机器增加到多台，执行部署命令。
1 ansible-playbook -i inventory/k8s/hosts.ini cluster.yml -b -vvv
例如：
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 kube-master-0 ansible_ssh_host=172.16.94.140 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.140 mask=/24 kube-master-1 ansible_ssh_host=172.16.94.144 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.144 mask=/24 kube-master-2 ansible_ssh_host=172.16.94.145 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.145 mask=/24 kube-node-41 ansible_ssh_host=172.16.94.141 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.141 mask=/24 kube-node-42 ansible_ssh_host=172.16.94.142 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.142 mask=/24 kube-node-43 ansible_ssh_host=172.16.94.143 ansible_ssh_user=root ansible_ssh_pass=123 ip=172.16.94.143 mask=/24 [kube-master] kube-master-0 kube-master-1 kube-master-2 [etcd] kube-master-0 kube-master-1 kube-master-2 [kube-node] kube-node-41 kube-node-42 kube-node-43 [k8s-cluster:children] kube-node kube-master [calico-rr]
6. 升级k8s集群
选择对应的k8s版本信息，执行升级命令。涉及文件为upgrade-cluster.yml。
1 ansible-playbook upgrade-cluster.yml -b -i inventory/k8s/hosts.ini -e kube_version=v1.10.4 -vvv
7. troubles shooting
在使用kubespary部署k8s集群时，主要遇到以下报错。
7.1. python-netaddr未安装
1 fatal: [node1]: FAILED! => {"failed" : true , "msg" : "The ipaddr filter requires python-netaddr be installed on the ansible controller" }
需要安装 python-netaddr，具体参考上述[环境准备]内容。
7.2. swap未关闭
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 fatal: [kube-master-0]: FAILED! => { "assertion" : "ansible_swaptotal_mb == 0" , "changed" : false , "evaluated_to" : false } fatal: [kube-node-41]: FAILED! => { "assertion" : "ansible_swaptotal_mb == 0" , "changed" : false , "evaluated_to" : false } fatal: [kube-node-42]: FAILED! => { "assertion" : "ansible_swaptotal_mb == 0" , "changed" : false , "evaluated_to" : false }
所有部署机器执行swapoff -a关闭swap，具体参考上述[环境准备]内容。
7.3. 部署机器内存过小
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 TASK [kubernetes/preinstall : Stop if memory is too small for masters] ********************************************************************************************************************************************************************************************************* task path: /root/gopath/src/kubespray/roles/kubernetes/preinstall/tasks/verify-settings.yml:52 Friday 10 August 2018 21:50:26 +0800 (0:00:00.940) 0:01:14.088 ********* fatal: [kube-master-0]: FAILED! => { "assertion" : "ansible_memtotal_mb >= 1500" , "changed" : false , "evaluated_to" : false } TASK [kubernetes/preinstall : Stop if memory is too small for nodes] *********************************************************************************************************************************************************************************************************** task path: /root/gopath/src/kubespray/roles/kubernetes/preinstall/tasks/verify-settings.yml:58 Friday 10 August 2018 21:50:27 +0800 (0:00:00.570) 0:01:14.659 ********* fatal: [kube-node-41]: FAILED! => { "assertion" : "ansible_memtotal_mb >= 1024" , "changed" : false , "evaluated_to" : false } fatal: [kube-node-42]: FAILED! => { "assertion" : "ansible_memtotal_mb >= 1024" , "changed" : false , "evaluated_to" : false } to retry, use: --limit @/root/gopath/src/kubespray/cluster.retry
调大所有部署机器的内存，本示例中调整为3G或以上。
7.4. kube-scheduler组件运行失败
kube-scheduler组件运行失败，导致http://localhost:10251/healthz调用失败。
1 2 3 FAILED - RETRYING: Master | wait for kube-scheduler (1 retries left). FAILED - RETRYING: Master | wait for kube-scheduler (1 retries left). fatal: [node1]: FAILED! => {"attempts" : 60, "changed" : false , "content" : "" , "failed" : true , "msg" : "Status code was not [200]: Request failed: <urlopen error [Errno 111] Connection refused>" , "redirected" : false , "status" : -1, "url" : "http://localhost:10251/healthz" }
可能是内存不足导致，本示例中调大了部署机器的内存。
7.5. docker安装包冲突
1 2 3 4 5 6 7 8 9 10 11 12 13 failed: [k8s-node-1] (item={u'name' : u'docker-engine-1.13.1-1.el7.centos' }) => { "attempts" : 4, "changed" : false , ... "item" : { "name" : "docker-engine-1.13.1-1.el7.centos" }, "msg" : "Error: docker-ce-selinux conflicts with 2:container-selinux-2.66-1.el7.noarch\n" , "rc" : 1, "results" : [ "Loaded plugins: fastestmirror\nLoading mirror speeds from cached hostfile\n * elrepo: mirrors.tuna.tsinghua.edu.cn\n * epel: mirrors.tongji.edu.cn\nPackage docker-engine is obsoleted by docker-ce, trying to install docker-ce-17.03.2.ce-1.el7.centos.x86_64 instead\nResolving Dependencies\n--> Running transaction check\n---> Package docker-ce.x86_64 0:17.03.2.ce-1.el7.centos will be installed\n--> Processing Dependency: docker-ce-selinux >= 17.03.2.ce-1.el7.centos for package: docker-ce-17.03.2.ce-1.el7.centos.x86_64\n--> Processing Dependency: libltdl.so.7()(64bit) for package: docker-ce-17.03.2.ce-1.el7.centos.x86_64\n--> Running transaction check\n---> Package docker-ce-selinux.noarch 0:17.03.2.ce-1.el7.centos will be installed\n---> Package libtool-ltdl.x86_64 0:2.4.2-22.el7_3 will be installed\n--> Processing Conflict: docker-ce-selinux-17.03.2.ce-1.el7.centos.noarch conflicts docker-selinux\n--> Restarting Dependency Resolution with new changes.\n--> Running transaction check\n---> Package container-selinux.noarch 2:2.55-1.el7 will be updated\n---> Package container-selinux.noarch 2:2.66-1.el7 will be an update\n--> Processing Conflict: docker-ce-selinux-17.03.2.ce-1.el7.centos.noarch conflicts docker-selinux\n--> Finished Dependency Resolution\n You could try using --skip-broken to work around the problem\n You could try running: rpm -Va --nofiles --nodigest\n" ] }
卸载旧的docker版本，由kubespary自动安装。
1 2 3 4 5 6 7 8 9 10 sudo yum remove -y docker \ docker-client \ docker-client-latest \ docker-common \ docker-latest \ docker-latest-logrotate \ docker-logrotate \ docker-selinux \ docker-engine-selinux \ docker-engine
参考文章：
支付宝打赏
微信打赏