Helm实战案例二：在Kubernetes(k8s)上使用helm安装部署日志管理系统EFK

目录

• 一.系统环境
• 二.前言
• 三.日志管理系统EFK简介
• 四.helm安装EFK
  • 4.1 helm在线安装EFK
  • 4.2 helm离线安装EFK(推荐)
• 五.访问kibana
  • 5.1 数据分片
• 六.卸载EFK
• 七.总结

一.系统环境

本文主要基于Kubernetes1.21.9和Linux操作系统CentOS7.4。

服务器版本	EFK版本	docker软件版本	Kubernetes(k8s)集群版本	CPU架构
CentOS Linux release 7.4.1708 (Core)	7.9.1	Docker version 20.10.12	v1.21.9	x86_64

Kubernetes集群架构：k8scloude1作为master节点，k8scloude2，k8scloude3作为worker节点。

服务器	操作系统版本	CPU架构	进程	功能描述
k8scloude1/192.168.110.130	CentOS Linux release 7.4.1708 (Core)	x86_64	docker，kube-apiserver，etcd，kube-scheduler，kube-controller-manager，kubelet，kube-proxy，coredns，calico	k8s master节点
k8scloude2/192.168.110.129	CentOS Linux release 7.4.1708 (Core)	x86_64	docker，kubelet，kube-proxy，calico	k8s worker节点
k8scloude3/192.168.110.128	CentOS Linux release 7.4.1708 (Core)	x86_64	docker，kubelet，kube-proxy，calico	k8s worker节点

二.前言

随着应用程序数量和规模的增长，日志管理变得越来越重要。日志管理系统可以帮助我们收集、存储、分析和可视化日志数据，以便快速诊断问题、监控性能等。在Kubernetes上部署一个可靠的日志管理系统是非常必要的。

在Kubernetes(k8s)上使用helm安装部署日志管理系统EFK的前提是已经有一套可以正常运行的Kubernetes集群，关于Kubernetes(k8s)集群的安装部署，可以查看博客《Centos7 安装部署Kubernetes(k8s)集群》https://www.cnblogs.com/renshengdezheli/p/16686769.html。

三.日志管理系统EFK简介

EFK是Elasticsearch、Fluentd和Kibana的缩写，它们都是开源软件。这三个软件组合在一起可以形成一个完整的日志管理系统，其中：

• Elasticsearch：用于存储、索引和搜索大量的结构化和非结构化数据。
• Fluentd：用于采集、转换和发送日志数据到Elasticsearch。
• Kibana：用于可视化和分析从Elasticsearch中检索到的数据。

EFK具有以下优点：

• 开源且免费。
• 扩展性好，可以处理大量的日志数据。
• 支持自定义插件，可以方便地进行扩展和定制。
• 支持实时查询和分析，可以快速诊断和解决问题。
• 支持多种数据可视化方式，例如柱状图、折线图等。

日志管理系统EFK架构如下：

除了EFK，你肯定还听说过ELK，ELK是Elasticsearch、Logstash和Kibana的缩写，它们也都是开源软件。和EFK类似，ELK也可以形成一个完整的日志管理系统，其中：

• Elasticsearch：用于存储、索引和搜索大量的结构化和非结构化数据。
• Logstash：用于采集、转换和发送日志数据到Elasticsearch。
• Kibana：用于可视化和分析从Elasticsearch中检索到的数据。

EFK和ELK的主要区别在于：

• Fluentd相对于Logstash更加轻量级，而且具有更好的性能。
• Elasticsearch相对于Logstash更加强大，可以自行处理数据采集和转换的任务，因此不需要使用Logstash。
• EFK更加灵活和易于扩展，支持自定义插件，可以方便地进行定制和扩展。

四.helm安装EFK

本文将介绍如何使用Helm在Kubernetes上部署一个日志管理系统EFK。关于helm的详细用法可以查看博客《Kubernetes(k8s)包管理工具Helm：Helm包管理》。

现在没有应用被安装。

[root@k8scloude1 ~]# helm ls
NAME	NAMESPACE	REVISION	UPDATED	STATUS	CHART	APP VERSION

添加EFK官方 helm 源。

[root@k8scloude1 ~]# helm repo add elastic https://helm.elastic.co
"elastic" has been added to your repositories

查看helm现在的仓库源。

[root@k8scloude1 ~]# helm repo list
NAME      	URL                                      
azure     	http://mirror.azure.cn/kubernetes/charts/
ali       	https://apphub.aliyuncs.com              
myhelmrepo	http://192.168.110.129/mycharts          
elastic   	https://helm.elastic.co       

在kubernetes集群的所有节点，创建存放EFK文件的目录。

[root@k8scloude1 ~]# mkdir efk

[root@k8scloude1 ~]# cd efk/

[root@k8scloude2 ~]# mkdir efk

[root@k8scloude2 ~]# cd efk/

[root@k8scloude3 ~]# mkdir efk

[root@k8scloude3 ~]# cd efk/

4.1 helm在线安装EFK

查询helm仓库里所有可得的elasticsearch。

[root@k8scloude1 ~]# helm search repo elasticsearch
NAME                        	CHART VERSION	APP VERSION	DESCRIPTION                                       
ali/elasticsearch           	11.0.4       	7.6.0      	A highly scalable open-source full-text search ...
......
azure/fluentd-elasticsearch 	2.0.7        	2.3.2      	DEPRECATED! - A Fluentd Helm chart for Kubernet...
elastic/elasticsearch       	7.16.3       	7.16.3     	Official Elastic helm chart for Elasticsearch     
ali/apm-server              	2.1.5        	7.0.0      	The server receives data from the Elastic APM a...
ali/dmarc2logstash          	1.2.0        	1.0.3      	Provides a POP3-polled DMARC XML report injecto...
......
elastic/eck-operator        	2.0.0        	2.0.0      	A Helm chart for deploying the Elastic Cloud on...
elastic/eck-operator-crds   	2.0.0        	2.0.0      	A Helm chart for installing the ECK operator Cu...

下载elasticsearch7.16.3版本，--version=7.16.3 指定下载的版本。

[root@k8scloude1 ~]# helm pull elastic/elasticsearch --version=7.16.3

下载elasticsearch7.9.1版本。

[root@k8scloude1 ~]# helm pull elastic/elasticsearch --version=7.9.1

现在两个不同版本的包已经下载下来了。

[root@k8scloude1 ~]# ls elasticsearch-7.*
elasticsearch-7.16.3.tgz  elasticsearch-7.9.1.tgz

解压。

[root@k8scloude1 ~]# tar xf elasticsearch-7.16.3.tgz 

[root@k8scloude1 ~]# cd elasticsearch/

可以查看版本号。

[root@k8scloude1 elasticsearch]# ls
Chart.yaml  examples  Makefile  README.md  templates  values.yaml

[root@k8scloude1 elasticsearch]# cat Chart.yaml 
apiVersion: v1
appVersion: 7.16.3
description: Official Elastic helm chart for Elasticsearch
home: https://github.com/elastic/helm-charts
icon: https://helm.elastic.co/icons/elasticsearch.png
maintainers:
- email: helm-charts@elastic.co
  name: Elastic
name: elasticsearch
sources:
- https://github.com/elastic/elasticsearch
version: 7.16.3

可以对应的修改values.yaml文件，进行配置。

[root@k8scloude1 elasticsearch]# vim values.yaml 

安装elasticsearch。

[root@k8scloude1 ~]# helm install elasticsearch elasticsearch

类似的，像这样安装Fluentd和Kibana即可。

4.2 helm离线安装EFK(推荐)

因为EFK所需的镜像都比较大，所以提前把EFK的镜像包，helm包都下载好，使用的是7.9.1版本，tar是镜像包，tgz是helm包。

[root@k8scloude1 efk]# ls
elastic7.9.1.tar  elasticsearch-7.9.1.tgz  filebeat7.9.1.tar  filebeat-7.9.1.tgz  kibana7.9.1.tar  kibana-7.9.1.tgz  metricbeat7.9.1.tar  metricbeat-7.9.1.tgz

把EFK镜像包传到其他worker节点，需要加载镜像。

[root@k8scloude1 efk]# scp elastic7.9.1.tar filebeat7.9.1.tar kibana7.9.1.tar metricbeat7.9.1.tar k8scloude2:/root/efk/
   
[root@k8scloude1 efk]# scp elastic7.9.1.tar filebeat7.9.1.tar kibana7.9.1.tar metricbeat7.9.1.tar k8scloude3:/root/efk/

接下来加载EFK镜像，所有节点都操作。

[root@k8scloude1 efk]# for i in elastic7.9.1.tar ; do docker load -i elastic7.9.1.tar ; docker load  -i filebeat7.9.1.tar ; docker load -i kibana7.9.1.tar ; docker load -i metricbeat7.9.1.tar ; done
613be09ab3c0: Loading layer [==================================================>]  211.1MB/211.1MB
......
13cbc7f1710b: Loading layer [==================================================>]  496.1kB/496.1kB
Loaded image: docker.elastic.co/elasticsearch/elasticsearch:7.9.1
5f5944cf8fa3: Loading layer [==================================================>]  23.96MB/23.96MB
......
00b987e8c2be: Loading layer [==================================================>]  374.8kB/374.8kB
Loaded image: docker.elastic.co/beats/filebeat:7.9.1
d81c06fbe333: Loading layer [==================================================>]  31.33MB/31.33MB
......
e2e362ce9c59: Loading layer [==================================================>]  374.8kB/374.8kB
Loaded image: docker.elastic.co/kibana/kibana:7.9.1
6f13949e424b: Loading layer [==================================================>]  138.9MB/138.9MB
......
29e2f99d88c3: Loading layer [==================================================>]  374.8kB/374.8kB
Loaded image: docker.elastic.co/beats/metricbeat:7.9.1

[root@k8scloude1 efk]# docker images | egrep "elastic|filebeat|kibana|metricbeat"
docker.elastic.co/kibana/kibana                                   7.9.1       2f79360c362e   17 months ago   1.18GB
docker.elastic.co/elasticsearch/elasticsearch                     7.9.1       22008d6d4b2d   17 months ago   762MB
docker.elastic.co/beats/metricbeat                                7.9.1       0a9e2b2ee772   17 months ago   505MB
docker.elastic.co/beats/filebeat                                  7.9.1       a48d035fe91b   17 months ago   449MB


[root@k8scloude2 efk]# for i in elastic7.9.1.tar ; do docker load -i elastic7.9.1.tar ; docker load  -i filebeat7.9.1.tar ; docker load -i kibana7.9.1.tar ; docker load -i metricbeat7.9.1.tar ; done

[root@k8scloude2 efk]# docker images | egrep "elastic|filebeat|kibana|metricbeat"
docker.elastic.co/kibana/kibana                                              7.9.1                 2f79360c362e   17 months ago   1.18GB
docker.elastic.co/elasticsearch/elasticsearch                                7.9.1                 22008d6d4b2d   17 months ago   762MB
docker.elastic.co/beats/metricbeat                                           7.9.1                 0a9e2b2ee772   17 months ago   505MB
docker.elastic.co/beats/filebeat                                             7.9.1                 a48d035fe91b   17 months ago   449MB


[root@k8scloude3 efk]# for i in elastic7.9.1.tar ; do docker load -i elastic7.9.1.tar ; docker load  -i filebeat7.9.1.tar ; docker load -i kibana7.9.1.tar ; docker load -i metricbeat7.9.1.tar ; done

[root@k8scloude3 efk]# docker images | egrep "elastic|filebeat|kibana|metricbeat"
docker.elastic.co/kibana/kibana                                       7.9.1       2f79360c362e   17 months ago   1.18GB
docker.elastic.co/elasticsearch/elasticsearch                         7.9.1       22008d6d4b2d   17 months ago   762MB
docker.elastic.co/beats/metricbeat                                    7.9.1       0a9e2b2ee772   17 months ago   505MB
docker.elastic.co/beats/filebeat                                      7.9.1       a48d035fe91b   17 months ago   449MB

已经提前把所有的EFK helm包都下载好了。

[root@k8scloude1 efk]# ls
elasticsearch-7.9.1.tgz  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat-7.9.1.tgz

解压elasticsearch。

[root@k8scloude1 efk]# tar xf elasticsearch-7.9.1.tgz 

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat-7.9.1.tgz

[root@k8scloude1 efk]# cd elasticsearch/

[root@k8scloude1 elasticsearch]# ls
Chart.yaml  examples  Makefile  README.md  templates  values.yaml

修改elasticsearch的values.yaml配置文件。

[root@k8scloude1 elasticsearch]# vim values.yaml 

/*
#elasticsearch/values.yaml文件修改内容如下
#因为我们k8s集群只有两个worker节点 ，所以副本数修改为2
replicas: 2
#一个master节点
minimumMasterNodes: 1

#去除持久性储存
    volumes:
      - secret
      - configMap
     # - persistentVolumeClaim

persistence:
  enabled: false

*/

所需镜像版本要修改和自己下载的版本一致。

[root@k8scloude1 elasticsearch]# grep image values.yaml 
image: "docker.elastic.co/elasticsearch/elasticsearch"
imageTag: "7.9.1"
imagePullPolicy: "IfNotPresent"

安装elasticsearch。

[root@k8scloude1 elasticsearch]# helm install elasticsearch .
W0317 11:09:06.285376   35427 warnings.go:70] policy/v1beta1 PodDisruptionBudget is deprecated in v1.21+, unavailable in v1.25+; use policy/v1 PodDisruptionBudget
W0317 11:09:06.322279   35427 warnings.go:70] policy/v1beta1 PodDisruptionBudget is deprecated in v1.21+, unavailable in v1.25+; use policy/v1 PodDisruptionBudget
NAME: elasticsearch
LAST DEPLOYED: Thu Mar 17 11:09:05 2022
NAMESPACE: helm
STATUS: deployed
REVISION: 1
NOTES:
1. Watch all cluster members come up.
  $ kubectl get pods --namespace=helm -l app=elasticsearch-master -w
2. Test cluster health using Helm test.
  $ helm test elasticsearch --cleanup

elasticsearch已经安装好了。

[root@k8scloude1 elasticsearch]# helm ls
NAME         	NAMESPACE	REVISION	UPDATED                                	STATUS  	CHART              	APP VERSION
elasticsearch	helm     	1       	2022-03-17 11:09:05.996472752 +0800 CST	deployed	elasticsearch-7.9.1	7.9.1    

查看pod，对应的pod也启动起来了。

[root@k8scloude1 elasticsearch]# kubectl get pod
NAME                     READY   STATUS    RESTARTS   AGE
elasticsearch-master-0   1/1     Running   0          3m52s
elasticsearch-master-1   1/1     Running   0          3m52s

接下来安装filebeat，解压filebeat。

[root@k8scloude1 elasticsearch]# cd ../

[root@k8scloude1 efk]# pwd
/root/efk

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat-7.9.1.tgz

[root@k8scloude1 efk]# tar xf filebeat-7.9.1.tgz 

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat-7.9.1.tgz

所需镜像版本要修改一致，filebeat/values.yaml 文件除了镜像版本，其他的不需要修改。

[root@k8scloude1 efk]# cd filebeat/

[root@k8scloude1 filebeat]# ls
Chart.yaml  examples  Makefile  README.md  templates  values.yaml

[root@k8scloude1 filebeat]# grep image values.yaml 
#   image: busybox
#   image: busybox
image: "docker.elastic.co/beats/filebeat"
imageTag: "7.9.1"
imagePullPolicy: "IfNotPresent"
imagePullSecrets: []

安装filebeat。

[root@k8scloude1 filebeat]# cd ../

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat-7.9.1.tgz

[root@k8scloude1 efk]# helm install filebeat filebeat
W0317 11:18:05.080789   42502 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:18:05.082311   42502 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 11:18:05.120788   42502 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:18:05.126392   42502 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
NAME: filebeat
LAST DEPLOYED: Thu Mar 17 11:18:04 2022
NAMESPACE: helm
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Watch all containers come up.
  $ kubectl get pods --namespace=helm -l app=filebeat-filebeat -w

filebeat的pod也启动起来了。

[root@k8scloude1 efk]# kubectl get pods -o wide
NAME                      READY   STATUS    RESTARTS   AGE     IP               NODE         NOMINATED NODE   READINESS GATES
elasticsearch-master-0    1/1     Running   0          18m     10.244.112.159   k8scloude2   <none>           <none>
elasticsearch-master-1    1/1     Running   0          18m     10.244.251.246   k8scloude3   <none>           <none>
filebeat-filebeat-5wl4j   1/1     Running   0          9m57s   10.244.112.161   k8scloude2   <none>           <none>
filebeat-filebeat-lkxxc   1/1     Running   0          9m57s   10.244.251.247   k8scloude3   <none>           <none>

解压metricbeat。

[root@k8scloude1 efk]# tar xf metricbeat-7.9.1.tgz 

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat  metricbeat-7.9.1.tgz

修改values.yaml文件，所需镜像版本要修改一致。

[root@k8scloude1 efk]# cd metricbeat/

[root@k8scloude1 metricbeat]# ls
charts  Chart.yaml  examples  Makefile  README.md  requirements.lock  requirements.yaml  templates  values.yaml

[root@k8scloude1 metricbeat]# grep image values.yaml
  # - Filesystem group for the metricbeat user. The official elastic docker images always have an id of 1000.
#   image: busybox
#   image: busybox
image: "docker.elastic.co/beats/metricbeat"
imageTag: "7.9.1"
imagePullPolicy: "IfNotPresent"
imagePullSecrets: []

安装metricbeat。

[root@k8scloude1 metricbeat]# cd ../

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat  metricbeat-7.9.1.tgz

#安装metricbeat
[root@k8scloude1 efk]# helm install metricbeat metricbeat
W0317 11:31:18.599946   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:31:18.601822   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:31:18.603664   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 11:31:18.605592   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 11:31:18.661010   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:31:18.661128   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 11:31:18.669323   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 11:31:18.671969   52640 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
NAME: metricbeat
LAST DEPLOYED: Thu Mar 17 11:31:18 2022
NAMESPACE: helm
STATUS: deployed
REVISION: 1
TEST SUITE: None
NOTES:
1. Watch all containers come up.
  $ kubectl get pods --namespace=helm -l app=metricbeat-metricbeat -w

metricbeat的pod启动起来了。

[root@k8scloude1 efk]# kubectl get pod -o wide
NAME                                             READY   STATUS    RESTARTS   AGE   IP               NODE         NOMINATED NODE   READINESS GATES
elasticsearch-master-0                           1/1     Running   0          23m   10.244.112.159   k8scloude2   <none>           <none>
elasticsearch-master-1                           1/1     Running   0          23m   10.244.251.246   k8scloude3   <none>           <none>
filebeat-filebeat-5wl4j                          1/1     Running   0          14m   10.244.112.161   k8scloude2   <none>           <none>
filebeat-filebeat-lkxxc                          1/1     Running   0          14m   10.244.251.247   k8scloude3   <none>           <none>
metricbeat-kube-state-metrics-75c5fc65d9-jcwz2   1/1     Running   0          55s   10.244.251.216   k8scloude3   <none>           <none>
metricbeat-metricbeat-metrics-7c5cd7d77f-tr4r4   1/1     Running   0          55s   10.244.112.166   k8scloude2   <none>           <none>
metricbeat-metricbeat-n62rf                      1/1     Running   0          55s   10.244.251.245   k8scloude3   <none>           <none>
metricbeat-metricbeat-w9m4j                      1/1     Running   0          55s   10.244.112.157   k8scloude2   <none>           <none>

解压kibana。

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana-7.9.1.tgz  metricbeat  metricbeat-7.9.1.tgz

[root@k8scloude1 efk]# tar xf kibana-7.9.1.tgz 

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana  kibana-7.9.1.tgz  metricbeat  metricbeat-7.9.1.tgz

把kibana/values.yaml里镜像版本修改一致。

[root@k8scloude1 efk]# grep image kibana/values.yaml 
image: "docker.elastic.co/kibana/kibana"
imageTag: "7.9.1"
imagePullPolicy: "IfNotPresent"
#   image: busybox
#   image: busybox
imagePullSecrets: []

kibana/values.yaml里，service服务的类型为ClusterIP，外界访问不了kibana的web界面，为了方便访问kibana，修改服务的发布类型为NodePort ，修改之后如下：

[root@k8scloude1 efk]# vim kibana/values.yaml 

[root@k8scloude1 efk]# egrep -A10 'service:' kibana/values.yaml
service:
  type: NodePort
  loadBalancerIP: ""
  port: 5601
  nodePort: ""
  labels: {}
  annotations: {}
    # cloud.google.com/load-balancer-type: "Internal"
    # service.beta.kubernetes.io/aws-load-balancer-internal: 0.0.0.0/0
    # service.beta.kubernetes.io/azure-load-balancer-internal: "true"
    # service.beta.kubernetes.io/openstack-internal-load-balancer: "true"

安装kibana。

[root@k8scloude1 efk]# ls
elasticsearch  elasticsearch-7.9.1.tgz  filebeat  filebeat-7.9.1.tgz  kibana  kibana-7.9.1.tgz  metricbeat  metricbeat-7.9.1.tgz

[root@k8scloude1 efk]# helm install kibana kibana
NAME: kibana
LAST DEPLOYED: Thu Mar 17 11:41:50 2022
NAMESPACE: helm
STATUS: deployed
REVISION: 1
TEST SUITE: None

查看pod，此时发现kibana-kibana-f88767f86-x4xpk一直处于Pending状态。

[root@k8scloude1 efk]# kubectl get pod -o wide 
NAME                                             READY   STATUS    RESTARTS   AGE   IP               NODE         NOMINATED NODE   READINESS GATES
elasticsearch-master-0                           1/1     Running   0          33m   10.244.112.159   k8scloude2   <none>           <none>
elasticsearch-master-1                           1/1     Running   0          33m   10.244.251.246   k8scloude3   <none>           <none>
filebeat-filebeat-5wl4j                          1/1     Running   0          24m   10.244.112.161   k8scloude2   <none>           <none>
filebeat-filebeat-lkxxc                          1/1     Running   0          24m   10.244.251.247   k8scloude3   <none>           <none>
kibana-kibana-f88767f86-x4xpk                    0/1     Pending   0          64s   <none>           <none>       <none>           <none>
metricbeat-kube-state-metrics-75c5fc65d9-jcwz2   1/1     Running   0          11m   10.244.251.216   k8scloude3   <none>           <none>
metricbeat-metricbeat-metrics-7c5cd7d77f-tr4r4   1/1     Running   0          11m   10.244.112.166   k8scloude2   <none>           <none>
metricbeat-metricbeat-n62rf                      1/1     Running   0          11m   10.244.251.245   k8scloude3   <none>           <none>
metricbeat-metricbeat-w9m4j                      1/1     Running   0          11m   10.244.112.157   k8scloude2   <none>           <none>

查找问题原因，发现两个原因：

• 原因1是：1 node(s) had taint {node-role.kubernetes.io/master，k8s集群的master节点默认有污点，不能把pod调度到master上；
• 原因2是：2 Insufficient cpu, 2 Insufficient memory. 说明内存CPU不够了。

[root@k8scloude1 ~]# kubectl describe pod kibana-kibana-f88767f86-x4xpk -n helm
Name:           kibana-kibana-f88767f86-x4xpk
Namespace:      helm
Priority:       0
Node:           <none>
......
    ConfigMapOptional:       <nil>
    DownwardAPI:             true
QoS Class:                   Guaranteed
Node-Selectors:              <none>
Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
Events:
  Type     Reason            Age    From               Message
  ----     ------            ----   ----               -------
  Warning  FailedScheduling  7m37s  default-scheduler  0/3 nodes are available: 1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate, 2 Insufficient cpu, 2 Insufficient memory.
  Warning  FailedScheduling  7m35s  default-scheduler  0/3 nodes are available: 1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate, 2 Insufficient cpu, 2 Insufficient memory.

对于node-role.kubernetes.io/master是污点这个问题，可以把污点去除，使pod能调度到master上，生产环境不推荐这么做，我这里机器配置不多就可以去除master上的污点。关于污点更详细内容请查看博客《pod(九)：污点taint 与容忍度tolerations》。

查看所有的污点。

[root@k8scloude1 ~]# kubectl get node -o yaml | grep taint -A 5
    taints:
    - effect: NoSchedule
      key: node-role.kubernetes.io/master
  status:
    addresses:
    - address: 192.168.110.130
--
      taint: T
    name: k8scloude2
    resourceVersion: "2653824"
    selfLink: /api/v1/nodes/k8scloude2
    uid: 618e3885-9918-4c29-b94c-1e074d40fa97
  spec:

清除master的污点。

[root@k8scloude1 ~]# kubectl taint nodes --all node-role.kubernetes.io/master-
node/k8scloude1 untainted
taint "node-role.kubernetes.io/master" not found
taint "node-role.kubernetes.io/master" not found

[root@k8scloude1 ~]# kubectl get node -o yaml | grep taint -A 5
      taint: T
    name: k8scloude2
    resourceVersion: "2654290"
    selfLink: /api/v1/nodes/k8scloude2
    uid: 618e3885-9918-4c29-b94c-1e074d40fa97
  spec:

再次查看kibana的pod描述信息，发现污点已经不报错了，现在只剩内存CPU不足报错：2 Insufficient memory, 3 Insufficient cpu。

[root@k8scloude1 ~]# kubectl describe pod kibana-kibana-f88767f86-x4xpk -n helm
Name:           kibana-kibana-f88767f86-x4xpk
Namespace:      helm
Priority:       0
Node:           <none>
......
    DownwardAPI:             true
QoS Class:                   Guaranteed
Node-Selectors:              <none>
Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
Events:
  Type     Reason            Age   From               Message
  ----     ------            ----  ----               -------
  Warning  FailedScheduling  21m   default-scheduler  0/3 nodes are available: 1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate, 2 Insufficient cpu, 2 Insufficient memory.
  Warning  FailedScheduling  21m   default-scheduler  0/3 nodes are available: 1 node(s) had taint {node-role.kubernetes.io/master: }, that the pod didn't tolerate, 2 Insufficient cpu, 2 Insufficient memory.
  Warning  FailedScheduling  29s   default-scheduler  0/3 nodes are available: 2 Insufficient memory, 3 Insufficient cpu.

查看k8s的负载。

[root@k8scloude1 efk]# kubectl top nodes
W0317 12:23:49.197137  107750 top_node.go:119] Using json format to get metrics. Next release will switch to protocol-buffers, switch early by passing --use-protocol-buffers flag
NAME         CPU(cores)   CPU%   MEMORY(bytes)   MEMORY%   
k8scloude1   251m         12%    1717Mi          52%       
k8scloude2   236m         11%    2485Mi          76%       
k8scloude3   206m         10%    2535Mi          78%       

CPU内存不足，最直接的解决办法就是机器加CPU和内存，还可以修改values.yaml 里的资源需求,我把EFK里的所有资源需求都列出来了。

/*
elasticsearch 	资源要求
resources:
  requests:
    cpu: "1000m"
    memory: "2Gi"
  limits:
    cpu: "1000m"
    memory: "2Gi"

filebeat  资源要求 
resources:
  requests:
    cpu: "100m"
    memory: "100Mi"
  limits:
    cpu: "1000m"
    memory: "200Mi"

kibana资源要求
resources:
  requests:
    cpu: "1000m"
    memory: "2Gi"
  limits:
    cpu: "1000m"
    memory: "2Gi"
        
metricbeat   资源要求
  resources:
    requests:
      cpu: "100m"
      memory: "100Mi"
    limits:
      cpu: "1000m"
      memory: "200Mi"
  tolerations: []
*/

我选择直接添加CPU和内存，重启机器添加CPU和内存之后，查看pod，可以看到EFK的pod都慢慢启动了。

[root@k8scloude1 ~]# kubectl get pod -o wide
NAME                                             READY   STATUS    RESTARTS   AGE     IP               NODE         NOMINATED NODE   READINESS GATES
elasticsearch-master-0                           0/1     Running   1          4h13m   10.244.112.165   k8scloude2   <none>           <none>
elasticsearch-master-1                           0/1     Running   1          4h13m   10.244.251.244   k8scloude3   <none>           <none>
filebeat-filebeat-4gqpl                          0/1     Running   1          3h21m   10.244.158.83    k8scloude1   <none>           <none>
filebeat-filebeat-5wl4j                          0/1     Running   1          4h4m    10.244.112.164   k8scloude2   <none>           <none>
filebeat-filebeat-lkxxc                          0/1     Running   1          4h4m    10.244.251.243   k8scloude3   <none>           <none>
kibana-kibana-f88767f86-x4xpk                    0/1     Running   0          3h40m   10.244.158.82    k8scloude1   <none>           <none>
metricbeat-kube-state-metrics-75c5fc65d9-jcwz2   1/1     Running   1          3h51m   10.244.251.215   k8scloude3   <none>           <none>
metricbeat-metricbeat-metrics-7c5cd7d77f-tr4r4   0/1     Running   1          3h51m   10.244.112.162   k8scloude2   <none>           <none>
metricbeat-metricbeat-n62rf                      0/1     Running   1          3h51m   10.244.251.252   k8scloude3   <none>           <none>
metricbeat-metricbeat-nsxwc                      0/1     Running   1          3h21m   10.244.158.79    k8scloude1   <none>           <none>
metricbeat-metricbeat-w9m4j                      0/1     Running   1          3h51m   10.244.112.163   k8scloude2   <none>           <none>

继续查看kibana的描述信息，查看pod的最新时间，可以看到kibana容器创建成功了，最新的告警Readiness probe failed: Error: Got HTTP code 000 but expected a 200，这是readiness probe探测警告，这个不影响，关于livenessprobe探测和readinessprobe探测详细内容，请查看博客《Kubernetes(k8s)健康性检查：livenessprobe探测和readinessprobe探测 》。

[root@k8scloude1 ~]# kubectl describe pod kibana-kibana-f88767f86-x4xpk -n helm
Name:         kibana-kibana-f88767f86-x4xpk
Namespace:    helm
Priority:     0
Node:         k8scloude1/192.168.110.130
Start Time:   Thu, 17 Mar 2022 15:22:01 +0800
.....
Node-Selectors:              <none>
Tolerations:                 node.kubernetes.io/not-ready:NoExecute op=Exists for 300s
                             node.kubernetes.io/unreachable:NoExecute op=Exists for 300s
Events:
  Type     Reason            Age                  From               Message
  ----     ------            ----                 ----               -------
  Warning  FailedScheduling  3h24m                default-scheduler  0/3 nodes are available: 2 Insufficient memory, 3 Insufficient cpu.
  Normal   Scheduled         5m21s                default-scheduler  Successfully assigned helm/kibana-kibana-f88767f86-x4xpk to k8scloude1
  Normal   Pulled            5m19s                kubelet            Container image "docker.elastic.co/kibana/kibana:7.9.1" already present on machine
  Normal   Created           5m19s                kubelet            Created container kibana
  Normal   Started           5m19s                kubelet            Started container kibana
  Warning  Unhealthy         91s (x22 over 5m1s)  kubelet            Readiness probe failed: Error: Got HTTP code 000 but expected a 200

现在所有的EFK pod都正常工作了。

[root@k8scloude1 ~]# kubectl get pod -o wide
NAME                                             READY   STATUS    RESTARTS   AGE     IP               NODE         NOMINATED NODE   READINESS GATES
elasticsearch-master-0                           1/1     Running   1          4h18m   10.244.112.165   k8scloude2   <none>           <none>
elasticsearch-master-1                           1/1     Running   1          4h18m   10.244.251.244   k8scloude3   <none>           <none>
filebeat-filebeat-4gqpl                          1/1     Running   1          3h26m   10.244.158.83    k8scloude1   <none>           <none>
filebeat-filebeat-5wl4j                          1/1     Running   1          4h9m    10.244.112.164   k8scloude2   <none>           <none>
filebeat-filebeat-lkxxc                          1/1     Running   1          4h9m    10.244.251.243   k8scloude3   <none>           <none>
kibana-kibana-f88767f86-x4xpk                    1/1     Running   0          3h46m   10.244.158.82    k8scloude1   <none>           <none>
metricbeat-kube-state-metrics-75c5fc65d9-jcwz2   1/1     Running   1          3h56m   10.244.251.215   k8scloude3   <none>           <none>
metricbeat-metricbeat-metrics-7c5cd7d77f-tr4r4   1/1     Running   1          3h56m   10.244.112.162   k8scloude2   <none>           <none>
metricbeat-metricbeat-n62rf                      1/1     Running   1          3h56m   10.244.251.252   k8scloude3   <none>           <none>
metricbeat-metricbeat-nsxwc                      1/1     Running   1          3h26m   10.244.158.79    k8scloude1   <none>           <none>
metricbeat-metricbeat-w9m4j                      1/1     Running   1          3h56m   10.244.112.163   k8scloude2   <none>           <none>

查看svc，kibana的外部访问端口是32053。

[root@k8scloude1 ~]# kubectl get svc
NAME                            TYPE        CLUSTER-IP     EXTERNAL-IP   PORT(S)             AGE
elasticsearch-master            ClusterIP   10.101.55.37   <none>        9200/TCP,9300/TCP   4h19m
elasticsearch-master-headless   ClusterIP   None           <none>        9200/TCP,9300/TCP   4h19m
kibana-kibana                   NodePort    10.103.1.29    <none>        5601:32053/TCP      3h46m
metricbeat-kube-state-metrics   ClusterIP   10.101.50.89   <none>        8080/TCP            3h57m

五.访问kibana

浏览器访问 http://192.168.110.130:32053/即可查看kibana的web界面。选择Explore on my own导入我们的数据。

kibana首页如下：

5.1 数据分片

下面进行数据分片，选择Home-->Discover。

点击Create index pattern。

如下为Create index pattern界面。

能看到如下四个index说明EFK环境没问题。

选择filebeat，点击下一步。

创建索引分区。

下面是filebeat*的所有字段。

再次点击Home-->Discover。

现在就可以看到日志了。

继续Create index pattern。

使用相同的方法创建index，选择metricbeat*。

点击Create index pattern。

metricbeat*的字段如下。

index patterns默认是filebeat*。

点击Discover。

可以切换查看不同的分区，右边显示的是对应的日志。

自此EFK搭建完毕。

六.卸载EFK

查看helm安装的应用。

[root@k8scloude1 ~]# helm ls
NAME         	NAMESPACE	REVISION	UPDATED                                	STATUS  	CHART              	APP VERSION
elasticsearch	helm     	1       	2022-03-17 11:09:05.996472752 +0800 CST	deployed	elasticsearch-7.9.1	7.9.1      
filebeat     	helm     	1       	2022-03-17 11:18:04.921145429 +0800 CST	deployed	filebeat-7.9.1     	7.9.1      
kibana       	helm     	1       	2022-03-17 11:41:50.305787166 +0800 CST	deployed	kibana-7.9.1       	7.9.1      
metricbeat   	helm     	1       	2022-03-17 11:31:18.353579301 +0800 CST	deployed	metricbeat-7.9.1   	7.9.1      

helm删除EFK。

[root@k8scloude1 ~]# helm delete elasticsearch
W0317 16:06:49.291827   76335 warnings.go:70] policy/v1beta1 PodDisruptionBudget is deprecated in v1.21+, unavailable in v1.25+; use policy/v1 PodDisruptionBudget
release "elasticsearch" uninstalled

[root@k8scloude1 ~]# helm delete filebeat
W0317 16:07:01.497539   76726 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 16:07:01.507858   76726 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
release "filebeat" uninstalled

[root@k8scloude1 ~]# helm delete kibana
release "kibana" uninstalled

[root@k8scloude1 ~]# helm delete metricbeat
W0317 16:07:14.955659   77279 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 16:07:14.955676   77279 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRoleBinding is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRoleBinding
W0317 16:07:15.014182   77279 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
W0317 16:07:15.014449   77279 warnings.go:70] rbac.authorization.k8s.io/v1beta1 ClusterRole is deprecated in v1.17+, unavailable in v1.22+; use rbac.authorization.k8s.io/v1 ClusterRole
release "metricbeat" uninstalled

删除EFK之后，对应的pod和svc都被删除了。

[root@k8scloude1 ~]# helm ls
NAME	NAMESPACE	REVISION	UPDATED	STATUS	CHART	APP VERSION

[root@k8scloude1 ~]# kubectl get pod
No resources found in helm namespace.

[root@k8scloude1 ~]# kubectl get svc
No resources found in helm namespace.

七.总结

通过本文，我们学会了如何使用Helm在Kubernetes上部署一个日志管理系统EFK。我们使用Helm安装了Elasticsearch、Fluentd和Kibana，访问了Kibana的Web界面，进行数据分片，最后看到了源源不断的日志。

本文给出的例子仅仅是EFK日志管理系统的一个基本配置，实际上可以根据需求进行进一步的定制和扩展。例如，需要更高级的数据可视化功能，可以将Grafana集成到EFK日志管理系统中；另外，在部署EFK时需要注意一些安全性问题，比如需要在Elasticsearch中设置访问控制等。