02 k8s考试基础知识（一）

k8s 有啥特点

• 服务发现&负载平衡， 服务很方便的给外部用户
• 方便回滚和故障恢复
• 有金主爸爸们（google 红帽之类的）

架构

• master（或者叫做Control Plane） 这边4个
  • etcl 存储这个分布式集群的信息
  • apiserver 通信用的
  • controller manage 这个是apiserver的小弟，没有主意的古惑仔
  • schedule 这个是apiserver的小弟之二，负责资源的调度
• worker 那边3个
  • kubelet 这个是apiserver的小弟之三，工具人
  • kube-proxy 这个复杂服务发现的比较重要，没有他，服务发现这个功能就寄了
  • container runtime 工具人

cka 模拟考题一

一个新的worker03节点，如何加入集群

# 1 进入master节点
ssh master
# 2 创建加入的命令
kubeadm
kubeadm token  -h 
kubeadm token  create -h 
kubeadm token  create --print-join-command
# 3 把第三个命令的输出结果记住，再到worker03节点执行
exit
ssh worker03
# 类似于下面的这个
kubeadm join 192.168.240.231:6443 --token xz51cy.bn85l1lh3xaargy8 --discovery-token-ca-cert-hash sha256:30502614fdcc0043f0cd2a4720b051f55feecdfd4db5ffee1b11250adead22b4
# 4 等5分钟
# 检验
kubectl get nodes

cka 模拟考题二

查看这个集群使用的是什么网络插件

# /etc/cni/net.d/ 这个目录下面显示啥，基本上就是啥了
ll /etc/cni/net.d/
# 05-cilium.conflist
k get po --all-namespaces | grep cilium
k get po -A | grep cilium

插一嘴 Pod QoS

pod总共有下面三种等级，这个不同的等级会在集群或者节点资源不充足的时候展现，展现的形式就是低等级的pod会最先被删除

• Guaranteed ：每个container中都设置了resources ，斌且其中的limits与requests 中包含cpu与memory的设置，并且一致
• Burstable：至少一个container中resources中有设置，设置成啥都随便
• BestEffort：一个container也没有设置resources

k explain ns
k explain ns.spec

# 创建实验环境
cat << eof | k apply -f - -n csdn-test
---
{ apiVersion: v1, kind: Namespace, metadata: { name: csdn-test } }
...
eof

for podd in `k get po -oname`; do k delete $podd ; done

• cpu/memory limit == cpu/memory request Guaranteed

• cpu/memory limits Guaranteed

• cpu/memory request Burstable

• cpu limits Burstable

• memory limits Burstable

• memory request Burstable

• cpu request Burstable

• BestEffort

所有container requests 啥也没设置 为 BestEffort

但凡有设置为Burstable

当所有containers中 limits 全设置 与 requests 与 limits 设置相等 则为 Guaranteed

cat << eof | k apply -f - -n csdn-test
---
apiVersion: v1
kind: Pod
metadata:
  name: p0
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      requests: { cpu: 200m, memory: 200Mi }
      limits: { cpu: 200m, memory: 200Mi }
---
apiVersion: v1
kind: Pod
metadata:
  name: p1
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
---
apiVersion: v1
kind: Pod
metadata:
  name: p2
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      requests: { cpu: 200m, memory: 200Mi }
---
apiVersion: v1
kind: Pod
metadata:
  name: p3
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 100m }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p4
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { memory: 200Mi }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p5
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      requests: { cpu: 100m }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p6
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      requests: { memory: 200Mi }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p7
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
...
eof

len=$(k get po -oname | wc -l)
for i in `seq $len`; do
  QoS=$(k describe po/p${i} | awk  '/QoS/{print $3}')
  printf '%s\t%s\n' $i $QoS
done
# k describe po/p1 |  sed -ne 's/QoS Class\:\s*//gp'

for podd in `k get po -oname`; do
  k delete $podd & 
done
cat << eof | k apply -f - -n csdn-test
---
apiVersion: v1
kind: Pod
metadata:
  name: p1
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
  - name: c2
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p2
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
  - name: c2
    image: nginx
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p3
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
      requests: { cpu: 200m, memory: 200Mi }
  - name: c2
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
      requests: { cpu: 200m, memory: 200Mi }
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p4
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
      requests: { cpu: 200m, memory: 200Mi }
  - name: c2
    image: nginx
...
---
apiVersion: v1
kind: Pod
metadata:
  name: p4
  namespace: csdn-test
spec:
  containers:
  - name: c1
    image: nginx
    resources:
      limits: { cpu: 200m, memory: 200Mi }
      requests: { cpu: 200m, memory: 200Mi }
  - name: c2
    image: nginx
...
eof

插一嘴 创建资源

# pod
k run po -h
# 直接运行
k run po --labels="aa=bb,bb=cc" --image=nginx --namespace=csdn-test --restart=Never  --image-pull-policy='Always' --port=80 --expose=true --command -- /bin/sh  \sleep 3600 ; 
# 编辑yaml
k run po   --labels="aa=bb,bb=cc" --image=nginx --namespace=csdn-test --restart=Never  --image-pull-policy='Always' --port=80 --expose=true --dry-run=client -oyaml 

插一嘴 静态pod

ssh master
ls /etc/kubernetes/manifests -l 

-rw-------. 1 root root 2445 12月 2 20:31 etcd.yaml
-rw------- 1 root root 3445 12月 24 00:29 kube-apiserver.yaml
-rw-------. 1 root root 2901 12月 2 20:31 kube-controller-manager.yaml
-rw-------. 1 root root 1487 12月 2 20:31 kube-scheduler.yaml

上面的这4个就是静态pod

k get po -n kube-system --no-headers
# 静态pod的名称后缀是-controle plane 主机的名称

这个路径是通过如下的方式获得的/etc/kubernetes/manifests

systemctl status kubelet
cd /usr/lib/systemd/system/kubelet.service.d
cat 10-kubeadm.conf 
cat /var/lib/kubelet/config.yaml | grep static

创建静态pod

k run sta-po   --labels="aa=bb,bb=cc" --image=nginx --namespace=csdn-test --restart=Never  --image-pull-policy='Always' --port=80  --dry-run=client -oyaml > /etc/kubernetes/manifests/aa.yml 
k get po -l "aa in (bb), bb in (cc)" -L aa,bb

NAME READY STATUS RESTARTS AGE AA BB
sta-po-huang-aa.xx 1/1 Running 0 70s bb cc