KubeSphere排错实战(二)

接上一篇《KubeSphere实战排除》近期在使用kubesphere中也记录了一些使用问题，希望可以对其他人有帮助

八 kubesphere应用上传问题

8.1 文件上传413

将应用部署进入kubesphere中，应用中有设置上传文件功能，测试上次异常无法正常上传，文件上传，ingress413报错，kubesphere使用的是ingress-nginx控制器，可以在其中注解添加k-v来支持，

解决方案：应用路由自定义max body size

https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#custom-max-body-size

8.2 大文件上传后端504

大文件上传后端响应504解决方案：

proxy read timeoutnginx.ingress.kubernetes.io/proxy-read-timeout

九 跨域问题

kubesphere使用ingress-nginx支持跨域，可以参考以下链接在注解中添加

https://kubernetes.github.io/ingress-nginx/user-guide/nginx-configuration/annotations/#enable-cors

测试环境可以使用可以使用hosts，将域名解析到本地，前端利用nginx来做静态文件服务，反向代理后端api，可以参考示例：

server {
  listen 80;
  server_name localhost;
  # 强制https跳转
  # rewrite ^(.*)$ https://$host$1 permanent;
  location / {
    index      index.html;
    root       /smart-frontend;
    try_files $uri $uri/ /index.html;
    client_body_buffer_size 200m;
    charset utf-8;
  }
  location /api {
      proxy_pass http://smart-backend:8080/api;
      proxy_read_timeout 1200;
      client_max_body_size 1024m;
  }
      gzip  on; #开启gzip
      gzip_vary on;
      gzip_min_length 1k; #不压缩临界值,大于1k的才压缩,一般不用改
      gzip_buffers 4 16k;
      gzip_comp_level 6; #压缩级别,数字越大压缩的越好
      gzip_types  text/plain application/javascript application/x-javascript text/css application/xml text/javascript application/x-httpd-php image/jpeg image/gif image/png image/x-icon;
}

十 添加节点

后期逐渐业务上来，集群节点资源不足，新增node节点，将node节点的数据盘添加到ceph节点

10.1 ceph集群添加节点

• 系统配置

• 免费密钥配置
• hosts配置
• docker安装并迁移至数据盘
• cgroup启用
• ceph数据节点添加

ceph集群配置添加node03集群的数据盘节点（如果数据存储类足够，可以不用添加数据节点）

[root@node03 docker]# mkfs.xfs /dev/vdd
[root@node03 docker]# mkdir -p /var/local/osd3
[root@node03 docker]# mount /dev/vdd /var/local/osd3/

添加vdd到/etc/fstab中
[root@node03 docker]# yum -y install yum-plugin-priorities epel-release

[root@node03 yum.repos.d]# chmod 777 -R /var/local/osd3/
[root@node03 yum.repos.d]# chmod 777 -R /var/local/osd3/*  master节点利用ceph-deploy部署node03节点[root@master ceph]# ceph-deploy install node03
[root@master ceph]# ceph-deploy  gatherkeys master
[root@master ceph]# ceph-deploy osd prepare node03:/var/local/osd3

• 激活osd

[root@master ceph]# ceph-deploy osd activate node03:/var/local/osd3

• 查看状态

[root@master ceph]# ceph-deploy osd list master node01 node02 node03

• 拷贝密钥

[root@master ceph]# ceph-deploy admin master node01 node02 node03

• 在node03节点设置权限

[root@node03 yum.repos.d]# chmod +r /etc/ceph/ceph.client.admin.keyring

• 在master设置MDS

[root@master ceph]# ceph-deploy mds create node01 node02 node03

• 查看状态

[root@master ceph]# ceph health
[root@master ceph]# ceph - 由于是新增node节点，数据需要平衡回填，此刻查看集群状态[root@master conf]# ceph -s
    cluster 5b9eb8d2-1c12-4f6d-ae9c-85078795794b
     health HEALTH_ERR
            44 pgs backfill_wait
            1 pgs backfilling
            1 pgs inconsistent
            45 pgs stuck unclean
            recovery 1/55692 objects degraded (0.002%)
            recovery 9756/55692 objects misplaced (17.518%)
            2 scrub errors
     monmap e1: 1 mons at {master=172.16.60.2:6789/0}
            election epoch 35, quorum 0 master
     osdmap e2234: 4 osds: 4 up, 4 in; 45 remapped pgs
            flags sortbitwise,require_jewel_osds
      pgmap v5721471: 192 pgs, 2 pools, 104 GB data, 27846 objects
            230 GB used, 1768 GB / 1999 GB avail
            1/55692 objects degraded (0.002%)
            9756/55692 objects misplaced (17.518%)
                 146 active+clean
                  44 active+remapped+wait_backfill
                   1 active+remapped+backfilling
                   1 active+clean+inconsistent
recovery io 50492 kB/s, 13 objects/s
  client io 20315 B/s wr, 0 op/s rd, 5 op/s wr 

• 最终的问题，目前由于新增了node节点，新增ceph数据节点需要数据同步

[root@master conf]# ceph -s
    cluster 5b9eb8d2-1c12-4f6d-ae9c-85078795794b
     health HEALTH_ERR
            1 pgs inconsistent
            2 scrub errors
     monmap e1: 1 mons at {master=172.16.60.2:6789/0}
            election epoch 35, quorum 0 master
     osdmap e2324: 4 osds: 4 up, 4 in
            flags sortbitwise,require_jewel_osds
      pgmap v5723479: 192 pgs, 2 pools, 104 GB data, 27848 objects
            229 GB used, 1769 GB / 1999 GB avail
                 191 active+clean
                   1 active+clean+inconsistent
  client io 78305 B/s wr, 0 op/s rd, 18 op/s wr修复[root@master conf]# ceph -s
    cluster 5b9eb8d2-1c12-4f6d-ae9c-85078795794b
     health HEALTH_OK
     monmap e1: 1 mons at {master=172.16.60.2:6789/0}
            election epoch 35, quorum 0 master
     osdmap e2324: 4 osds: 4 up, 4 in
            flags sortbitwise,require_jewel_osds
      pgmap v5724320: 192 pgs, 2 pools, 104 GB data, 27848 objects
            229 GB used, 1769 GB / 1999 GB avail
                 192 active+clean
  client io 227 kB/s wr, 0 op/s rd, 7 op/s wr
# 同步完成
[root@master conf]# ceph health
HEALTH_OK

10.2 node节点添加

kubesphere为方便新增节点，提供了方便的脚步一键新增，可参考：https://kubesphere.com.cn/docs/v2.1/zh-CN/installation/add-nodes/

修改host.ini

[all]
master ansible_connection=local  ip=172.16.60.2
node01  ansible_host=172.16.60.3  ip=172.16.60.3 
node02  ansible_host=172.16.60.4  ip=172.16.60.4
node03  ansible_host=172.16.60.5  ip=172.16.60.5
[kube-master]
master           
[kube-node]
master
node01   
node02
node03

在 "/script" 目录执行 add-nodes.sh脚本。待扩容脚本执行成功后，即可看到包含新节点的集群节点信息，可通过 KubeSphere 控制台的菜单选择 基础设施 然后进入 主机管理 页面查看，或者通过 Kubectl 工具执行 kubectl get node命令，查看扩容后的集群节点详细信息。

[root@master scripts]# ./add-nodes.sh

查看验证

[root@master conf]# kubectl get nodes -owide
NAME     STATUS   ROLES         AGE    VERSION   INTERNAL-IP   EXTERNAL-IP   OS-IMAGE                KERNEL-VERSION          CONTAINER-RUNTIME
master   Ready    master        136d   v1.15.5   172.16.60.2           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node01   Ready    node,worker   136d   v1.15.5   172.16.60.3           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node02   Ready    node,worker   136d   v1.15.5   172.16.60.4           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node03   Ready    worker        10m    v1.15.5   172.16.60.5           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://19.3.5
[root@master conf]# kubectl label node  node-role.kubernetes.io/node=
common.yaml            hosts.ini              plugin-qingcloud.yaml  
[root@master conf]# kubectl label node node03  node-role.kubernetes.io/node=   
node/node03 labeled
[root@master conf]# kubectl get nodes -owide                                
NAME     STATUS   ROLES         AGE    VERSION   INTERNAL-IP   EXTERNAL-IP   OS-IMAGE                KERNEL-VERSION          CONTAINER-RUNTIME
master   Ready    master        136d   v1.15.5   172.16.60.2           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node01   Ready    node,worker   136d   v1.15.5   172.16.60.3           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node02   Ready    node,worker   136d   v1.15.5   172.16.60.4           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://18.6.2
node03   Ready    node,worker   11m    v1.15.5   172.16.60.5           CentOS Linux 7 (Core)   3.10.0-693.el7.x86_64   docker://19.3.5
[root@master conf]# 

十一 K8s集群资源不均

可以发现k8s资源使用不均衡，之前的部署应用为制定nodeSelect，导致一些系统服务运行在node节点，查看node2内存占用很大,导致集群异常告警或重启

可以通过查看

kubectl get pods -o wide --all-namespaces |grep node02 |awk '{print $1,  $2}'

将一些系统应用通过nodeselect来调度到master节点，以减轻node2节点的内存压力。

`kubectl  get nodes --show-labels`

在node2上查看系统组建添加nodeselector来重新调度

      nodeSelector:
        node-role.kubernetes.io/master: master

查看现存在node2上面的kubesphere系统deployment

通过调度完成，查看node2的内存负载已经下来了

十二 kubesphere devops工程

新增了node03节点，devops工程一周为队列中为此时运行该job的实例未完成初始化，登录集群查看，node03上的base pod在pull agent镜像，为了快速，直接在node节点，save base镜像然后在node03上load

[root@master ~]# kubectl describe pods -n kubesphere-devops-system $(kubectl get pods -n kubesphere-devops-system |grep -E "^base" |awk '{print $1}')

十三 kubesphere 应用安装

目前自己的kubesphere集群为2.1，在具体的在项目中添加了repo后，后台回自己去同步镜像数据还是需要为手动在那个地方触发下，我添加了几个helm 的repo，好像里面的chart没有在web界面显示，在添加了repo的项目下，我新建应用，然后选择来自kubesphere的应用商店，其中只有几个charts，发现不了添加的helm 源的chartscharts，在服务器内部是可以使用命令search到。咨询社区暂时为收到回复，记得v2.0版本后台有个任务会去同步charts，目前2.1版本，先使用helm命令在集群内进行手动helm安装

[root@master common-service]# helm install -n consul --namespace common-service -f consul/values-production.yaml consul/
NAME:   consul
LAST DEPLOYED: Tue Jan 14 17:56:27 2020
NAMESPACE: common-service
STATUS: DEPLOYED

RESOURCES:
==> v1/Pod(related)
NAME      READY  STATUS   RESTARTS  AGE
consul-0  0/2    Pending  0         0s

==> v1/Service
NAME       TYPE       CLUSTER-IP   EXTERNAL-IP  PORT(S)                                                AGE
consul     ClusterIP  None                8400/TCP,8301/TCP,8301/UDP,8300/TCP,8600/TCP,8600/UDP  1s
consul-ui  ClusterIP  10.233.59.7         80/TCP                                                 1s
==> v1/StatefulSet
NAME    READY  AGE
consul  0/3    0s

==> v1beta1/PodDisruptionBudget
NAME        MIN AVAILABLE  MAX UNAVAILABLE  ALLOWED DISRUPTIONS  AGE
consul-pdb  1              N/A              0                    1s

NOTES:
  ** Please be patient while the chart is being deployed **

  Consul can be accessed within the cluster on port 8300 at consul.common-service.svc.cluster.local

In order to access to the Consul Web UI:

    kubectl port-forward --namespace common-service svc/consul-ui 80:80
    echo "Consul URL: http://127.0.0.1:80"

Please take into account that you need to wait until a cluster leader is elected before using the Consul Web UI.

In order to check the status of the cluster you can run the following command:

    kubectl exec -it consul-0 -- consul members

Furthermore, to know which Consul node is the cluster leader run this other command:

    kubectl exec -it consul-0 -- consul operator raf

具体问题可以参考帖子：https://kubesphere.com.cn/forum/d/669-kubesphere

自己整理了k8s学习笔记，有兴起的可以一快学习交流：https://github.com/redhatxl/awesome-kubernetes-notes
支持国产容器管理平台KubeSphere,为社区尽自己的一份绵薄之力。