【运维知识大神篇】运维界的超神器Kubernetes教程2（资源清单编写+镜像下载策略+容器重启策略+资源限制详解+端口映射详解+K8s故障排查技巧）
2024-01-09 11:51:48
本篇文章继续给大家介绍Kubernetes，主要是Pod的进阶知识，包括资源清单的编写，镜像下载策略，容器重启策略，资源限制详解，端口映射详解，五大K8s故障排查技巧。带你深入了解Pod的魅力。
目录
资源清单编写
要擅于使用 kubectl explain 查看文档，及官方文档的查看
环境变量传递，一个资源清单运行两个项目，可看上一篇文章的实战项目
镜像下载策略
镜像下载策略，指的是利用k8s启动容器时，是如何下载镜像的，有三种策略。
Always：总是取拉去最新的镜像，本地有的话会覆盖；
Never：如果本地有镜像，则会用本地镜像，尝试启动容器；如果本地没有镜像，则永远不会去拉取尝试镜像；
IfNotPresent：如果本地有镜像，本地镜像存在同名称的tag，会取出镜像的RepoDigests（镜像摘要）和远程仓库的RepoDigests进行比较，若比较结果相同，则直接使用本地缓存镜像，若比较结果不同，则会拉去远程仓库的最新镜像。
当镜像标签是latest时，默认策略是Always；当镜像标签是自定义时（也就是标签不是latest），那么默认策略是IfNotPresent。
我们准备在仓库提前上传一个镜像，在本地再准备一个同名镜像，测试修改资源清单的配置文件判断是否在仓库下载了镜像
1、worker232打包镜像，远程上传linux1，linux2目录都创建的，本地删除，再打包个只创建linux1目录的
#编写要上传到仓库的dockerfile
[root@Master232 dockerfile]# cat dockerfile
FROM alpine:latest
RUN mkdir /koten-linux1
RUN mkdir /koten-linux2
CMD ["tail","-f","/etc/hosts"]
#构建dockerfile
[root@Master232 dockerfile]# docker build -t harbor.koten.com/koten-linux/alpine:latest .
#上传dockerfile到私有仓库
[root@Master232 dockerfile]# docker push harbor.koten.com/koten-linux/alpine:latest
#删除本地镜像，此时本地没有镜像
[root@Master232 dockerfile]# docker rmi harbor.koten.com/koten-linux/alpine:latest
Untagged: harbor.koten.com/koten-linux/alpine:latest
Untagged: harbor.koten.com/koten-linux/alpine@sha256:64e6642cba7dae211c954797d59e42368dcb3e12ee03d4a443660382548b827c
Deleted: sha256:46b6cef27ac0c2db9e67e6242986f4bf5eb229cc94adece6071ae9792985c6bf
Deleted: sha256:9838d1c75c25086c17bdfd0f7ae5d8b7fe10260797065e0e6670b6c7f4de3e37
Deleted: sha256:ebddcc1cca4e837f6c4a3242d887d07aae17266cd02a41333571a504bc726bb6
#修改dockerfile为只创建一个目录
[root@Master232 dockerfile]# cat dockerfile
FROM alpine:latest
RUN mkdir /koten-linux1
CMD ["tail","-f","/etc/hosts"]
#再次构建镜像
[root@Master232 dockerfile]# docker build -t harbor.koten.com/koten-linux/alpine:latest .2、编写资源清单，测试镜像
此时的仓库镜像为创建两个目录的，而本地镜像为创建一个目录的，我们通过资源清单运行
测试IfNotPresent下载策略
[root@Master231 pod]# cat 05-pods-download.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-pull-policy
spec:
# nodeName: worker233
nodeName: worker232
containers:
- name: image-pull-policy
image: harbor.koten.com/koten-linux/alpine:latest
# 镜像的拉取策略
# imagePullPolicy: Always
# imagePullPolicy: Never
imagePullPolicy: IfNotPresent
[root@Master231 pod]# kubectl apply -f 05-pods-download.yaml
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-pull-policy 1/1 Running 0 47s
#发现只有linux1，本地有镜像所以不再下载
[root@Master231 pod]# kubectl exec -it image-pull-policy -- ls /
bin koten-linux1 opt sbin usr
dev lib proc srv var
etc media root sys
home mnt run tmp
?测试Always下载策略
[root@Master231 pod]# cat 05-pods-download.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-pull-policy
spec:
# nodeName: worker233
nodeName: worker232
containers:
- name: db
image: harbor.koten.com/koten-linux/alpine:latest
# 镜像的拉取策略
imagePullPolicy: Always
# imagePullPolicy: Never
# imagePullPolicy: IfNotPresent
[root@Master231 pod]# kubectl delete pods --all
[root@Master231 pod]# kubectl apply -f 05-pods-download.yaml
pod/image-pull-policy created
#发现有两个目录，所以确定是在远端下载运行了
[root@Master231 pod]# kubectl exec -it image-pull-policy -- ls /
bin koten-linux1 mnt run tmp
dev koten-linux2 opt sbin usr
etc lib proc srv var
home media root sys测试Never下载策略
[root@Master231 pod]# cat 05-pods-download.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-pull-policy
spec:
# nodeName: worker233
nodeName: worker232
containers:
- name: db
image: harbor.koten.com/koten-linux/alpine:latest
# 镜像的拉取策略
# imagePullPolicy: Always
imagePullPolicy: Never
# imagePullPolicy: IfNotPresent
[root@Master231 pod]# kubectl apply -f 05-pods-download.yaml
pod/image-pull-policy created
#选择Never，刚刚本地被覆盖了，已经有创建两个目录的镜像了，所以采用本地镜像运行
[root@Master231 pod]# kubectl exec -it image-pull-policy -- ls /
bin koten-linux1 mnt run tmp
dev koten-linux2 opt sbin usr
etc lib proc srv var
home media root sys
#删除镜像再次运行资源清单
[root@Worker232 dockerfile]# docker image rm -f harbor.koten.com/koten-linux/alpine:latest
Untagged: harbor.koten.com/koten-linux/alpine:latest
Untagged: harbor.koten.com/koten-linux/alpine@sha256:a5b9fbd47d6897edcf1f6a8d6df13bbc1005c54241256069653499c53e66bc2c
#发现并没有去寻找镜像，直接报错
[root@Master231 pod]# kubectl apply -f 05-pods-download.yaml
pod/image-pull-policy created
[root@Master231 pod]# kubectl exec -it image-pull-policy -- ls /
error: unable to upgrade connection: container not found ("db")容器重启策略
容器退出时，容器是否重新创建新的容器，有三个参数。
Always：容器退出时，始终重启容器（创建新容器），是默认重启策略
Never：容器退出时，不重启容器（不创建新容器）
OnFailure：当容器异常退出时（kill -9时容器的退出码不是0，是137），重启容器（即创建容器），当容器正常退出（docker stop，退出码为0），不重启容器。
当Pod中的容器退出时，kubelet会按照指数回退方式计算重启的延迟（10s、20s、40s......），最长延迟为5分钟，如果容器重启后执行10分钟没有出现问题，kubelet会重启这个延迟时间。
注意：
1、无论容器的重启策略是什么，当我们手动使用它docker移除容器时，K8S均会自动拉起并不会记录重启次数;
2、当容器非正常退出(即异常退出，可以使用kill -9模拟)时，Always和OnFailure这两种策略会重新拉起POD并会记录重启次数;
3、当任务正常退出时(docker stop、docker rm)，只有Always可以重启任务并记录重启次数，但是Always有时候也记录不上正常退出的，关停后重启了，但是次数没有加;
4、kubectl delete pods --all删除Pod不会触发重启策略。编写资源清单依次按照三个重启策略运行
[root@Master231 pod]# cat 06-pods-restartPolicy.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-restart-policy-01
spec:
nodeName: worker232
# 指定容器的重启策略
restartPolicy: Always
# restartPolicy: Never
# restartPolicy: OnFailure
containers:
- name: db
image: harbor.koten.com/koten-linux/alpine:latest
imagePullPolicy: Always
# 修改容器的运行命令
command: ["sleep","60"]
测试Always重启策略
[root@Master231 pod]# kubectl apply -f 06-pods-restartPolicy.yaml
pod/image-restart-policy-01 created
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-restart-policy-01 1/1 Running 0 6s
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
91da1ebbe971 harbor.koten.com/koten-linux/alpine "sleep 60" 18 seconds ago Up 17 seconds k8s_db_image-restart-policy-01_default_428e644e-f9c4-41a2-8460-afd16bf8d5bf_0
#用docker停止容器，但是立即重启
[root@Worker232 dockerfile]# docker stop 91da1ebbe971
91da1ebbe971
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
4c5515f0c9ed harbor.koten.com/koten-linux/alpine "sleep 60" 4 seconds ago Up 3 seconds k8s_db_image-restart-policy-01_default_428e644e-f9c4-41a2-8460-afd16bf8d5bf_1
[root@Worker232 dockerfile]# ps -axu|grep sleep
root 74726 0.0 0.0 1568 252 ? Ss 20:36 0:00 sleep 60
root 74902 0.0 0.0 112808 968 pts/0 S+ 20:37 0:00 grep --color=auto sleep
[root@Worker232 dockerfile]# kill -9 74726
[root@Worker232 dockerfile]# ps -axu|grep sleep
root 75004 0.0 0.0 112808 968 pts/0 S+ 20:37 0:00 grep --color=auto sleep
#等了一会儿，又出现了
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
15ccf66844ca harbor.koten.com/koten-linux/alpine "sleep 60" 9 seconds ago Up 8 seconds k8s_db_image-restart-policy-01_default_428e644e-f9c4-41a2-8460-afd16bf8d5bf_2
#主节点会记录重启次数
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-restart-policy-01 0/1 CrashLoopBackOff 2 (20s ago) 2m59s测试Never重启策略
[root@Master231 pod]# cat 06-pods-restartPolicy.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-restart-policy-01
spec:
nodeName: worker232
# 指定容器的重启策略
# restartPolicy: Always
restartPolicy: Never
# restartPolicy: OnFailure
containers:
- name: db
image: harbor.koten.com/koten-linux/alpine:latest
imagePullPolicy: Always
# 修改容器的运行命令
command: ["sleep","60"]
[root@Master231 pod]# kubectl delete pods --all
pod "image-restart-policy-01" deleted
[root@Master231 pod]# kubectl apply -f 06-pods-restartPolicy.yaml
pod/image-restart-policy-01 created
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
e817061abc25 harbor.koten.com/koten-linux/alpine "sleep 60" 12 seconds ago Up 10 seconds k8s_db_image-restart-policy-01_default_0c9b4e80-3fd7-475a-85c0-9ddca90b7860_0
[root@Worker232 dockerfile]# docker stop e817061abc25
#停止后Pod状态变成Error（容器异常退出），不会再次启动
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-restart-policy-01 0/1 Error 0 58s测试OnFailure重启策略
[root@Master231 pod]# cat 06-pods-restartPolicy.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-restart-policy-01
spec:
nodeName: worker232
# 指定容器的重启策略
# restartPolicy: Always
# restartPolicy: Never
restartPolicy: OnFailure
containers:
- name: db
image: harbor.koten.com/koten-linux/alpine:latest
imagePullPolicy: Always
# 修改容器的运行命令
command: ["sleep","60"]
[root@Master231 pod]# kubectl delete pods --all
pod "image-restart-policy-01" deleted
[root@Master231 pod]# kubectl apply -f 06-pods-restartPolicy.yaml
pod/image-restart-policy-01 created
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
c3d84673ec3a harbor.koten.com/koten-linux/alpine "sleep 60" 10 seconds ago Up 10 seconds k8s_db_image-restart-policy-01_default_80ba68c5-18b0-4529-9cd0-a8baa7bb370c_0
#异常退出后重启
[root@Worker232 dockerfile]# ps -axu|grep sleep
root 78357 0.0 0.0 1568 252 ? Ss 20:43 0:00 sleep 60
root 78710 0.0 0.0 112808 964 pts/0 S+ 20:44 0:00 grep --color=auto sleep
[root@Worker232 dockerfile]# kill -9 78357
[root@Worker232 dockerfile]# ps -axu|grep sleep
root 78813 0.0 0.0 112808 968 pts/0 S+ 20:44 0:00 grep --color=auto sleep
[root@Worker232 dockerfile]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
dce4aa08b442 harbor.koten.com/koten-linux/alpine "sleep 60" 32 seconds ago Up 32 seconds k8s_db_image-restart-policy-01_default_80ba68c5-18b0-4529-9cd0-a8baa7bb370c_2
#正常退出后不再重启
[root@Worker232 dockerfile]# docker stop dce4aa08b442
dce4aa08b442
#此时状态为Completed（容器正常退出）
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-restart-policy-01 0/1 Completed 2 4m6s资源限制详解
在资源清单上编写参数，用于限制Pod，Namespace的资源，比如Pod内的某个容器资源使用cpu，内存等。
拉取压测工具镜像
1、拉取测试镜像，打标签，发送至harbor私有仓库
先在harbor上创建名为koten-tools的项目
docker pull bosskoten/stress-tools:v1.0
docker tag bosskoten/stress-tools:v1.0 harbor.koten.com/koten-tools/stress:v0.1
docker push harbor.koten.com/koten-tools/stress:v0.12、编写资源清单
[root@Master231 pod]# cat 07-pods-resources.yaml
apiVersion: v1
kind: Pod
metadata:
name: image-resources-stress-04
spec:
# nodeName: worker232
restartPolicy: Always
containers:
- name: stress
# image: bosskoten/stress-tools:v1.0
image: harbor.koten.com/koten-tools/stress:v0.1
imagePullPolicy: Always
command: ["tail","-f","/etc/hosts"]
# 配置资源限制
resources:
# 指定容器期望资源，若满足不了，则无法完成调度，状态会显示异常。但是满足了也不一定能用到这些，只是根据业务需求来说，用这些合适
# 若超出资源限制，状态会显示为OutOfmemory（内存溢出异常）
requests:
# 指定内存限制
# memory: "2G"
memory: "200M"
# 1core=1000m
cpu: "250m"
# 指定资源的上限，超出则业务异常，可能是容器被终止，可能是告警，可能是节点负载增加，可能是业务性能下降
limits:
memory: "500M"
# 表示使用1.5core，即1500m
cpu: 1.5
#cpu: 0.53、运行资源清单
[root@Master231 pod]# kubectl apply -f 07-pods-resources.yaml
pod/image-resources-stress-04 created
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
image-resources-stress-04 1/1 Running 0 10s
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
image-resources-stress-04 1/1 Running 0 24s 10.100.1.10 worker232 <none> <none>
4、进行压力测试，观察是否超过限制
#启用资源清单进行测试
[root@Worker232 ~]# docker ps
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
1123a7e941a2 harbor.koten.com/koten-tools/stress "tail -f /etc/hosts" About a minute ago Up About a minute k8s_stress_image-resources-stress-04_default_7c870e99-76e7-42a1-807b-8b81bdca5645_0
#进行CPU压测，产生四个CPU进程1分钟后停止运行
[root@Master231 pod]# kubectl exec -it image-resources-stress-04 sh
/usr/local/stress # stress -c 4 --verbose --timeout 1m
#发现虽然指定了四颗CPU，但是CPU使用率稳定在150%左右
[root@Worker232 ~]# docker stats 1123a7e941a2
CONTAINER ID NAME CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O PIDS
1123a7e941a2 k8s_stress_image-resources-stress-04_default_7c870e99-76e7-42a1-807b-8b81bdca5645_0 149.25% 316KiB / 476.8MiB 0.06% 0B / 0B 1.52MB / 0B 7
#产生2个work工作经常，并且每个工作经常占用200000000 Bytes(即200MB)，且不释放内存，但不会超过400M的使用空间。
/usr/local/stress # stress -m 2 --vm-bytes 200000000 --vm-keep --verbose
#发现内存占用在400M左右，并且后面有大约500M的限制
CONTAINER ID NAME CPU % MEM USAGE / LIMIT MEM % NET I/O BLOCK I/O PIDS
1123a7e941a2 k8s_stress_image-resources-stress-04_default_7c870e99-76e7-42a1-807b-8b81bdca5645_0 150.69% 383.1MiB / 476.8MiB 80.34% 0B / 0B 4.39MB / 0B 5端口映射详解
K8s集群外部访问K8s的Pod业务有6种方式，分别是hostNetwork；hostIP；svc的nodePort类型；apiServerr的反向代理，也是svc的CluserIP；kube-proxy；ingress；
?上篇文章在实战项目中演示了了hostNetwork和端口映射，现在以wordpress举例，详细讲解下hostIP端口映射
1、先编写资源清单，端口映射可以合着写也可以分着容器写
[root@Master231 pod]# cat 08-pods-ports.yaml
apiVersion: v1
kind: Pod
metadata:
name: wordpress
spec:
#hostNetwork: true
nodeName: worker233
containers:
- name: db
image: harbor.koten.com/koten-db/mysql:5.7
# 向容器传递环境变量
env:
# 变量的名称
- name: MYSQL_ALLOW_EMPTY_PASSWORD
# 指定变量的值
value: "yes"
- name: MYSQL_DATABASE
value: "wordpress"
- name: MYSQL_USER
value: "admin"
- name: MYSQL_PASSWORD
value: "123"
ports:
# 容器的端口号
- containerPort: 3306
# 容器绑定到宿主机的IP地址，若不写表示"0.0.0.0"
hostIP: 10.0.0.233
# 映射的主机端口
hostPort: 13306
# 给服务起名字，要求唯一
name: db
# 指定协议
protocol: TCP
#- containerPort: 80
# hostIP: 10.0.0.233
# hostPort: 18080
# name: wp
- name: wordpress
image: harbor.koten.com/koten-web/wordpress:v1.0
env:
- name: WORDPRESS_DB_HOST
# 指定变量的值
value: "127.0.0.1"
- name: WORDPRESS_DB_NAME
value: "wordpress"
- name: WORDPRESS_DB_USER
value: "admin"
- name: WORDPRESS_DB_PASSWORD
value: "123"
ports:
- containerPort: 80
hostIP: 10.0.0.233
hostPort: 18080
name: wp
2、运行资源清单
[root@Master231 pod]# kubectl apply -f 08-pods-ports.yaml
pod/wordpress created
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
wordpress 2/2 Running 0 5s 10.100.2.9 worker233 <none> <none>
3、查看端口发现没有端口映射，防火墙中可以找到端口转发，直接curl IP也可以curl通
[root@Worker233 ~]# ss -ntl|grep 18080
[root@Worker233 ~]# iptables -t nat -vnL|grep 18080
0 0 CNI-HOSTPORT-SETMARK tcp -- * * 10.100.2.0/24 10.0.0.233 tcp dpt:18080
0 0 CNI-HOSTPORT-SETMARK tcp -- * * 127.0.0.1 10.0.0.233 tcp dpt:18080
0 0 DNAT tcp -- * * 0.0.0.0/0 10.0.0.233 tcp dpt:18080 to:10.100.2.9:80
0 0 CNI-DN-7391794342a07a3e211cd tcp -- * * 0.0.0.0/0 0.0.0.0/0 /* dnat name: "cbr0" id: "33057eacf048ab12986c65024b0928476c1d1473cd280bda7be9d50ef8321942" */ multiport dports 13306,18080
[root@Master231 pod]# curl -I 10.100.2.9
HTTP/1.1 302 Found
Date: Thu, 15 Jun 2023 09:13:18 GMT
Server: Apache/2.4.51 (Debian)
X-Powered-By: PHP/7.4.27
Expires: Wed, 11 Jan 1984 05:00:00 GMT
Cache-Control: no-cache, must-revalidate, max-age=0
X-Redirect-By: WordPress
Location: http://10.100.2.9/wp-admin/install.php
Content-Type: text/html; charset=UTF-8
4、将WordPress案例工作节点拆分，一个pod两个容器分别运行在两个节点，要求使用ports字段进行暴露，要求wordpres在232节点，数据库在233节点，不使用宿主机网络。
编写资源清单
[root@Master231 pod]# cat 09-pods-ports-split.yaml
apiVersion: v1
kind: Pod
metadata:
name: koten-mysql
spec:
nodeName: worker233
containers:
- name: mysql
image: harbor.koten.com/koten-db/mysql:5.7
# 向容器传递环境变量
env:
# 变量的名称
- name: MYSQL_ALLOW_EMPTY_PASSWORD
# 指定变量的值
value: "yes"
- name: MYSQL_DATABASE
value: "wordpress"
- name: MYSQL_USER
value: "admin"
- name: MYSQL_PASSWORD
value: "123"
ports:
- containerPort: 3306
name: db
---
apiVersion: v1
kind: Pod
metadata:
name: koten-wordpress
spec:
nodeName: worker232
containers:
- name: wordpress
image: harbor.koten.com/koten-web/wordpress:v1.0
env:
- name: WORDPRESS_DB_HOST
# 指定变量的值
value: "127.0.0.1"
- name: WORDPRESS_DB_NAME
value: "wordpress"
- name: WORDPRESS_DB_USER
value: "admin"
- name: WORDPRESS_DB_PASSWORD
value: "123"
ports:
- containerPort: 80
hostIP: 10.0.0.232
hostPort: 18080
name: wp
运行资源清单查看mysql的IP?
[root@Master231 pod]# kubectl apply -f 09-pods-ports-split.yaml
pod/koten-mysql created
pod/koten-wordpress created
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
koten-mysql 1/1 Running 0 40s 10.100.2.10 worker233 <none> <none>
koten-wordpress 1/1 Running 0 40s 10.100.1.32 worker232 <none> <none>
获取到IP后修改资源清单，只删除后wordpress的pod，再apply更新
[root@Master231 pod]# cat 09-pods-ports-split.yaml
......
- name: WORDPRESS_DB_HOST
# 指定变量的值
value: "10.100.2.10"
......
[root@Master231 pod]# kubectl delete pod koten-wordpress
pod "koten-wordpress" deleted
[root@Master231 pod]# kubectl apply -f 09-pods-ports-split.yaml
pod/koten-mysql unchanged
pod/koten-wordpress created
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
koten-mysql 1/1 Running 0 8m35s 10.100.2.10 worker233 <none> <none>
koten-wordpress 1/1 Running 0 17s 10.100.1.33 worker232 <none> <none>可以curl通，浏览器访问正常，部署正常
由于是80的端口映射在宿主机的18080端口上，所以我们curl容器的ip不带端口，默认80，浏览器访问宿主机的IP需要带18080端口
[root@Master231 pod]# curl -I 10.100.1.33
HTTP/1.1 302 Found
Date: Thu, 15 Jun 2023 13:22:14 GMT
Server: Apache/2.4.51 (Debian)
X-Powered-By: PHP/7.4.27
Expires: Wed, 11 Jan 1984 05:00:00 GMT
Cache-Control: no-cache, must-revalidate, max-age=0
X-Redirect-By: WordPress
Location: http://10.100.1.33/wp-admin/install.php
Content-Type: text/html; charset=UTF-8
直接跳转到了这一步说明传输的信息有效了?
K8s故障排查技巧
一、kubectl descrribe
查看资源的详细信息，根据事件信息获取当前资源的状态，从而给出解决方案
三种查看方式
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
koten-mysql 1/1 Running 0 14m
koten-wordpress 1/1 Running 0 5m43s
[root@Master231 pod]# kubectl describe -f 09-pods-ports-split.yaml
[root@Master231 pod]# kubectl describe pod koten-wordpress
[root@Master231 pod]# kubectl describe pod/koten-wordpress
......
Events:
Type Reason Age From Message
---- ------ ---- ---- -------
Normal Pulled 33m kubelet Container image "harbor.koten.com/koten-web/wordpress:v1.0" already present on machine
Normal Created 33m kubelet Created container wordpress
Normal Started 33m kubelet Started container wordpress
[root@Master231 pod]# 一般情况下我们查看里面的Events即可，里面有报错信息，如果Pod状态显示Pending，没有Events，可以观察详细信息的Node字段与节点信息的名称是否一致
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
koten-mysql 1/1 Running 0 17s 10.100.2.11 worker233 <none> <none>
koten-wordpress 0/1 Pending 0 17s <none> work232 <none> <none>
[root@Master231 pod]# kubectl describe pod koten-wordpress
[root@Master231 pod]# kubectl describe pod koten-wordpress
Name: koten-wordpress
Namespace: default
Priority: 0
Node: work232/
Labels: <none>
Annotations: <none>
......
Events: <none>
[root@Master231 pod]# kubectl get no
NAME STATUS ROLES AGE VERSION
master231 Ready control-plane,master 29h v1.23.17
worker232 Ready <none> 29h v1.23.17
worker233 Ready <none> 29h v1.23.17
二、kubectl logs
通过容器启动的日志查看
实时查看某个Pod的日志
[root@Master231 pod]# kubectl logs -f games
+ nginx
+ '[' -n ]
+ '[' -n ]
+ echo root:123
+ chpasswd
chpasswd: password for 'root' changed
+ ssh-keygen -A
+ /usr/sbin/sshd -D
查看Pod中指定容器的日志
[root@Master231 pod]# kubectl logs -f koten-wp -c wordpress查看Pod中某个容器最近2分钟内产生的日志
[root@Master231 pod]# kubectl logs --since=2m -f koten-wp -c wordpress查看之前终止的 Pod 的日志。如果 Pod 已经重新启动，它将查看最新一次终止的 Pod 的日志。前提是该容器未被删除
kubectl logs -p -f games三、kubectl exec
连接到指定的容器查看
编写一个Pod节点有两个容器的资源清单
[root@Master231 pod]# cat 09-pods-ports-split.yaml
apiVersion: v1
kind: Pod
metadata:
name: koten-wp
spec:
nodeName: worker233
containers:
- name: mysql
image: harbor.koten.com/koten-db/mysql:5.7
# 向容器传递环境变量
env:
# 变量的名称
- name: MYSQL_ALLOW_EMPTY_PASSWORD
# 指定变量的值
value: "yes"
- name: MYSQL_DATABASE
value: "wordpress"
- name: MYSQL_USER
value: "admin"
- name: MYSQL_PASSWORD
value: "123"
ports:
- containerPort: 3306
name: db
- name: wordpress
image: harbor.koten.com/koten-web/wordpress:v1.0
env:
- name: WORDPRESS_DB_HOST
# 指定变量的值
value: "127.0.0.1"
- name: WORDPRESS_DB_NAME
value: "wordpress"
- name: WORDPRESS_DB_USER
value: "admin"
- name: WORDPRESS_DB_PASSWORD
value: "123"
ports:
- containerPort: 80
hostIP: 10.0.0.232
hostPort: 18080
name: wp
运行资源清单，发现一个Pod显示两个容器，查询Pod
[root@Master231 pod]# kubectl apply -f 09-pods-ports-split.yaml
pod/koten-wp created
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
koten-wp 2/2 Running 0 7s
#当一个Pod只有一个容器时，可以直接进入
[root@Master231 pod]# kubectl exec -it koten-mysql -- sh
#当一个Pod有多个容器时，可以-c指定容器进入
[root@Master231 pod]# kubectl exec -it koten-wp -c wordpress -- bash
root@koten-wp:/var/www/html#
#若不指定，则默认进入资源清单里第一个容器
[root@Master231 pod]# kubectl exec -it koten-wp -- bash
Defaulted container "mysql" out of: mysql, wordpress
root@koten-wp:/#
四、kubectl get
-o? yaml：用于获取资源的资源清单，避免资源清单丢失无法找到的情况
-o? wide：以宽格式显示资源信息
观察输出字段，Ready判断容器是否就绪，Status观察容器运行的状态，Restart判断容器是否重启、重启次数
[root@Master231 pod]# kubectl get pods
NAME READY STATUS RESTARTS AGE
koten-wp 2/2 Running 0 5m38s
[root@Master231 pod]# kubectl get pods -o yaml
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
koten-wp 2/2 Running 0 6m50s 10.100.2.15 worker233 <none> <none>五、kubectl cp
用户拷贝数据
拷贝本地的/etc/hosts到pod中db容器的/haha路径
kubectl cp /etc/hosts koten-wp:/haha 将本地的目录拷贝到容器的指定路径
kubectl cp /etc koten-wp:/haha -c db将Pod的db容器的/haha目录拷贝到宿主机，更名为xixi目录
kubectl cp koten-wp:/haha xixi -c db将当前目录下的所有文件打包，并解压至指定Pod的容器的指定路径
tar cf - * | kubectl exec -i koten-wp -c db -- tar xf - -C /tmp/六、command和args区别
[root@Master231 pod]# cat 14-pods-command-args.yaml
apiVersion: v1
kind: Pod
metadata:
name: linux-command-args
spec:
# hostNetwork: true
containers:
- name: web
image: harbor.koten.com/koten-web/nginx:1.25.1-alpine
# 类似于Dockerfile的ENTRYPOINT指令。
# command: ["tail","-f","/etc/hosts"]
# 类似于Dockerfile的CMD指令。
# args: ["sleep","3600"]
command:
- "tail"
- "-f"
args:
- "/etc/hosts"
[root@Master231 pod]# kubectl apply -f 14-pods-command-args.yaml
pod/linux-command-args created
[root@Master231 pod]# kubectl get pods -o wide
NAME READY STATUS RESTARTS AGE IP NODE NOMINATED NODE READINESS GATES
linux-command-args 1/1 Running 0 53s 10.100.1.43 worker232 <none> <none>
查看镜像的命令
[root@Worker232 ~]# docker ps -l
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
71d7ce418ecd 4a20372d4940 "tail -f /etc/hosts" About a minute ago Up About a minute k8s_web_linux-command-args_default_1ad12102-4fe3-4e78-96e4-e560cd51dcd1_0
我是koten，10年运维经验，持续分享运维干货，感谢大家的阅读和关注
文章来源:https://blog.csdn.net/qq_37510195/article/details/131216851
本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。 如若内容造成侵权/违法违规/事实不符，请联系我的编程经验分享网邮箱：veading@qq.com进行投诉反馈，一经查实，立即删除！
本文来自互联网用户投稿，该文观点仅代表作者本人，不代表本站立场。本站仅提供信息存储空间服务，不拥有所有权，不承担相关法律责任。 如若内容造成侵权/违法违规/事实不符，请联系我的编程经验分享网邮箱：veading@qq.com进行投诉反馈，一经查实，立即删除！