Ubuntu下的K8s部署

2021-07-04 00:48:06

1、 K8S 的由来

K8S 是 kubernetes 的英文缩写，是用 8 代替 8 个字符 "ubernete" 而成的缩写。




2、 K8S 单机版实战

环境：
  • ubuntu 16.04
  • gpu 驱动 418.56
  • docker 18.06
  • k8s 1.13.5

一、设置环境

首先备份一下源配置:cp /etc/apt/sources.list /etc/apt/sources.list.cp

编辑vim /etc/apt/sources.list，替换为阿里源:

deb-src http://archive.ubuntu.com/ubuntu xenial main restricted
deb http://mirrors.aliyun.com/ubuntu/ xenial main restricted
deb-src http://mirrors.aliyun.com/ubuntu/ xenial main restricted multiverse universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-updates main restricted multiverse universe
deb http://mirrors.aliyun.com/ubuntu/ xenial universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates universe
deb http://mirrors.aliyun.com/ubuntu/ xenial multiverse
deb http://mirrors.aliyun.com/ubuntu/ xenial-updates multiverse
deb http://mirrors.aliyun.com/ubuntu/ xenial-backports main restricted universe multiverse
deb-src http://mirrors.aliyun.com/ubuntu/xenial-backports main restricted universe multiverse
deb http://archive.canonical.com/ubuntu xenial partner
deb-src http://archive.canonical.com/ubuntu xenial partner
deb http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted
deb-src http://mirrors.aliyun.com/ubuntu/ xenial-security main restricted multiverse universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-security universe
deb http://mirrors.aliyun.com/ubuntu/ xenial-security multiverse

更新源:apt-get update

自动修复安装出现 broken 的 package:apt --fix-broken install

升级，对于 gpu 机器可不执行，否则可能升级 gpu 驱动导致问题:apt-get upgrade

关闭防火墙:ufw disable

安装 selinux:apt install selinux-utils

selinux 防火墙配置:

setenforce 0

vim/etc/selinux/conifg

SELINUX=disabled

设置网络:

tee /etc/sysctl.d/k8s.conf <<-'EOF'
net.bridge.bridge-nf-call-ip6tables = 1
net.bridge.bridge-nf-call-iptables = 1
net.ipv4.ip_forward = 1
EOF

modprobe br_netfilter

查看 ipv4 与 v6 配置是否生效:sysctl --system

配置 iptables:

iptables -P FORWARD ACCEPT

vim /etc/rc.local
/usr/sbin/iptables -P FORWARD ACCEPT

永久关闭 swap 分区:sed -i 's/.*swap.*/#&/' /etc/fstab

二、安装 docker

执行下面的命令:

apt-get install apt-transport-https ca-certificates curl software-properties-common

curl -fsSL https://mirrors.aliyun.com/docker-ce/linux/ubuntu/gpg | apt-key add -

add-apt-repository "deb [arch=amd64] https://download.docker.com/linux/ubuntu $(lsb_release -cs) stable" apt-get update

apt-get purge docker-ce docker docker-engine docker.io  && rm -rf /var/lib/docker

apt-get autoremove docker-ce docker docker-engine docker.io

apt-get install -y docker-ce=18.06.3~ce~3-0~ubuntu

启动 docker 并设置开机自重启:systemctl enable docker && systemctl start docker

Docker 配置:

vim /etc/docker/daemon.json
{
 "log-driver": "json-file",
 "log-opts": {
   "max-size": "100m",
   "max-file": "10"
 },
 "insecure-registries": ["http://k8s.gcr.io"],
 "data-root": "",
 "default-runtime": "nvidia",
 "runtimes": {
     "nvidia": {
         "path": "/usr/bin/nvidia-container-runtime",
         "runtimeArgs": []
     }
 }
}

上面是含 GPU 的配置，不含 GPU 的配置:

{
"registry-mirrors":[
"https://registry.docker-cn.com"
],
"storage-driver":"overlay2",
"log-driver":"json-file",
"log-opts":{
"max-size":"100m"
},
"exec-opts":[
"native.cgroupdriver=systemd"
],
"insecure-registries":["http://k8s.gcr.io"],
"live-restore":true
}

重启服务并设置开机自动重启:

systemctl daemon-reload && systemctl restart docker && docker info

三、安装 k8s

拉取镜像前的设置:

apt-get update && apt-get install -y apt-transport-https curl

curl -s https://mirrors.aliyun.com/kubernetes/apt/doc/apt-key.gpg | apt-key add -

tee /etc/apt/sources.list.d/kubernetes.list <<-'EOF'
deb https://mirrors.aliyun.com/kubernetes/apt kubernetes-xenial main
EOF

更新:

apt-get update
apt-get purge kubelet=1.13.5-00 kubeadm=1.13.5-00 kubectl=1.13.5-00
apt-get autoremove kubelet=1.13.5-00 kubeadm=1.13.5-00 kubectl=1.13.5-00
apt-get install -y kubelet=1.13.5-00 kubeadm=1.13.5-00 kubectl=1.13.5-00
apt-mark hold kubelet=1.13.5-00 kubeadm=1.13.5-00 kubectl=1.13.5-00

启动服务并设置开机自动重启:

systemctl enable kubelet && sudo systemctl start kubelet

安装 k8s 相关镜像，由于 gcr.io 网络访问不了，从 registry.cn-hangzhou.aliyuncs.com 镜像地址下载:

docker pull registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-apiserver:v1.13.5

docker pull registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-controller-manager:v1.13.5

docker pull registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-scheduler:v1.13.5

docker pull registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-proxy:v1.13.5

docker pull registry.cn-hangzhou.aliyuncs.com/kuberimages/pause:3.1

docker pull registry.cn-hangzhou.aliyuncs.com/kuberimages/etcd:3.2.24

docker pull registry.cn-hangzhou.aliyuncs.com/kuberimages/coredns:1.2.6

打标签:

docker tag registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-apiserver:v1.13.5 k8s.gcr.io/kube-apiserver:v1.13.5

docker tag registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-controller-manager:v1.13.5 k8s.gcr.io/kube-controller-manager:v1.13.5

docker tag registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-scheduler:v1.13.5 k8s.gcr.io/kube-scheduler:v1.13.5

docker tag registry.cn-hangzhou.aliyuncs.com/gg-gcr-io/kube-proxy:v1.13.5 k8s.gcr.io/kube-proxy:v1.13.5

docker tag registry.cn-hangzhou.aliyuncs.com/kuberimages/pause:3.1 k8s.gcr.io/pause:3.1

docker tag registry.cn-hangzhou.aliyuncs.com/kuberimages/etcd:3.2.24 k8s.gcr.io/etcd:3.2.24

docker tag registry.cn-hangzhou.aliyuncs.com/kuberimages/coredns:1.2.6 k8s.gcr.io/coredns:1.2.6

四、kubeadm 初始化

利用 kubeadm 初始化 k8s，其中主机 IP 根据自己的实际情况输入:

kubeadm init --kubernetes-version=v1.13.5 --pod-network-cidr=10.244.0.0/16 --service-cidr=10.16.0.0/16 --apiserver-advertise-address=${masterIp} | tee kubeadm-init.log

此时，如果未知主机 IP，也可利用 yaml 文件动态初始化:

vi /etc/hosts
10.10.5.100 k8s.api.server

vi kube-init.yaml

apiVersion: kubeadm.k8s.io/v1beta1
kind: ClusterConfiguration
kubernetesVersion: v1.13.5
imageRepository: registry.aliyuncs.com/google_containers
apiServer:
  certSANs:
  - "k8s.api.server"
controlPlaneEndpoint: "k8s.api.server:6443"
networking:
  serviceSubnet: "10.1.0.0/16"
  podSubnet: "10.244.0.0/16"

HA 版本:

apiVersion: kubeadm.k8s.io/v1beta1
kind: ClusterConfiguration
kubernetesVersion: v1.13.5
imageRepository: registry.aliyuncs.com/google_containers
apiServer:
  certSANs:
  - "api.k8s.com"
controlPlaneEndpoint: "api.k8s.com:6443"
etcd:
    external:
        endpoints:
        - https://ETCD_0_IP:2379
        - https://ETCD_1_IP:2379
        - https://ETCD_2_IP:2379
networking:
  serviceSubnet: 10.1.0.0/16
  podSubnet: 10.244.0.0/16

注意: apiVersion 中用 kubeadm，因为需要用 kubeadm 来初始化，最后执行下面来初始化:

kubeadm init --config=kube-init.yaml

出现问题，解决后，reset 后再执行，如果需要更多，执行:

 kubeadm --help

五、部署出现问题

先删除 node 节点(集群版)

kubectl drain <node name> --delete-local-data --force --ignore-daemonsets
​
kubectl delete node <node name>

清空 init 配置在需要删除的节点上执行（注意，当执行 init 或者 join 后出现任何错误，都可以使用此命令返回）:

kubeadm reset

六、查问题

初始化后出现问题，可以通过以下命令先查看其容器状态以及网络情况:

sudo docker ps -a | grep kube | grep -v pause

sudo docker logs CONTAINERID

sudo docker images && systemctl status -l kubelet

netstat -nlpt

kubectl describe ep kubernetes

kubectl describe svc kubernetes

kubectl get svc kubernetes

kubectl get ep

netstat -nlpt | grep apiser

vi /var/log/syslog

七、给当前用户配置 k8s apiserver 访问公钥

sudo mkdir -p $HOME/.kube

sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config

sudo chown $(id -u):$(id -g) $HOME/.kube/config

八、网络插件

kubectl apply -f https://docs.projectcalico.org/v3.3/getting-started/kubernetes/installation/hosted/rbac-kdd.yaml

wget https://docs.projectcalico.org/v3.3/getting-started/kubernetes/installation/hosted/kubernetes-datastore/calico-networking/1.7/calico.yaml

vi calico.yaml

- name: CALICO_IPV4POOL_IPIP
 value:"off"
- name: CALICO_IPV4POOL_CIDR
 value: "10.244.0.0/16

kubectl apply -f calico.yaml

单机下允许 master 节点部署 pod 命令如下:

kubectl taint nodes --all node-role.kubernetes.io/master-

禁止 master 部署 pod:

kubectl taint nodes k8s node-role.kubernetes.io/master=true:NoSchedule

以上单机版部署结束，如果你的项目中，交付的是软硬件结合的一体机，那么到此就结束了。记得单机下要允许 master 节点部署哟！

接下来，集群版本上线咯！

以上面部署的机器为例，作为 master 节点，继续执行:

scp /etc/kubernetes/admin.conf $nodeUser@$nodeIp:/home/$nodeUser

scp /etc/kubernetes/pki/etcd/* $nodeUser@$nodeIp:/home/$nodeUser/etcd

kubeadm token generate

kubeadm token create $token_name --print-join-command --ttl=0

kubeadm join $masterIP:6443 --token  $token_name --discovery-token-ca-cert-hash $hash

Node 机器执行时，如果需要 cuda ，可以参考以下资料:

https://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html#ubuntu-installation
https://blog.csdn.net/u012235003/article/details/54575758
https://blog.csdn.net/qq_39670011/article/details/90404111

正式执行:

vim /etc/modprobe.d/blacklist-nouveau.conf
blacklist nouveau
options nouveau modeset=0
update-initramfs -u

重启 ubuntu 查看是否禁用成功:

lsmod | grep nouveau

apt-get remove --purge nvidia*

https://developer.nvidia.com/cuda-downloads

sudo apt-get install freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libgl1-mesa-glx libglu1-mesa libglu1-mesa-dev

安装 cuda:

accept
select "Install" / Enter
select "Yes"
sh cuda_10.1.168_418.67_linux.run

echo 'export PATH=/usr/local/cuda-10.1/bin:$PATH' >> ~/.bashrc
echo 'export PATH=/usr/local/cuda-10.1/NsightCompute-2019.3:$PATH' >> ~/.bashrc
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-10.1/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

重启机器，检查 cuda 是否安装成功。

查看是否有“nvidia*”的设备:

cd /dev && ls -al

如果没有，创建一个 nv.sh:

vi nv.sh
#!/bin/bash /sbin/modprobe nvidia
if [ "$?" -eq 0 ];
then
NVDEVS=`lspci |
 grep -i NVIDIA
`
N3D=`
echo
"$NVDEVS"
| grep "3D controller" |
 wc -l
`
NVGA=`
echo
"$NVDEVS"
| grep "VGA compatible controller" |
 wc -l
`
N=`
expr $N3D + $NVGA -
1
`
for i in `
seq
0
 $N
`; do
    mknod -m 666 /dev/nvidia$i c 195 $i
done
    mknod -m 666 /dev/nvidiactl c 195 255
else
    exit 1
fi

chmod +x nv.sh && bash nv.sh

再次重启机器查看 cuda 版本:

nvcc -V

编译:

cd /usr/local/cuda-10.1/samples &&  make

cd  /usr/local/cuda-10.1/samples/bin/x86_64/linux/release ./deviceQuery

以上如果输出：“Result = PASS” 代表 cuda 安装成功。

安装 nvdocker:

vim /etc/docker/daemon.json
{
"runtimes":{
    "nvidia":{
         "path":"nvidia-container-runtime",
          "runtimeArgs":[]
     }
},
"registry-mirrors":["https://registry.docker-cn.com"],
"storage-driver":"overlay2",
"default-runtime":"nvidia",
"log-driver":"json-file",
"log-opts":{
 "max-size":"100m"
},
"exec-opts": ["native.cgroupdriver=systemd"],
"insecure-registries": [$harborRgistry],
"live-restore": true
}

重启 docker:

sudo systemctl daemon-reload && sudo systemctl restart docker && docker info

检查 nvidia-docker 安装是否成功:

docker run --runtime=nvidia --rm nvidia/cuda:9.0-base nvidia-smi

在节点机器进入 su 模式:su $nodeUser

给当前节点用户配置 k8s apiserver 访问公钥:

mkdir -p $HOME/.kube

cp -i  admin.conf $HOME/.kube/config

chown $(id -u):$(id -g) $HOME/.kube/config

mkdir -p $HOME/etcd

sudo rm -rf /etc/kubernetes

sudo mkdir -p /etc/kubernetes/pki/etcd

sudo cp /home/$nodeUser/etcd/* /etc/kubernetes/pki/etcd

sudo kubeadm join $masterIP:6443 --token  $token_name --discovery-token-ca-cert-hash $hash

如:

sudo kubeadm join 192.168.8.116:6443 --token vyi4ga.foyxqr2iz9i391q3 --discovery-token-ca-cert-hash sha256:929143bcdaa3e23c6faf20bc51ef6a57df02edf9df86cedf200320a9b4d3220a

检查 node 是否加入 master:kubectl get node

以上介绍了单机的 k8s 部署，以及 HA 的 master 节点的部署安装。
上一篇：成功解决在Eclipse中出现的can not find declarration to go to


下一篇：Morsel-Driven Parallelism: A NUMA-Aware Query Evaluation Framework for the Many-Core Age 论文解读