image

前言

Elasticsearch + Logstash + Kibana（ELK）是一套開源的日志管理方案，分析網站的訪問情況時我們一般會借助Google/百度/CNZZ等方式嵌入JS做數據統計，但是當網站訪問異?；蛘弑还魰r我們需要在后臺分析如Nginx的具體日志，而Nginx日志分割/GoAccess/Awstats都是相對簡單的單節點解決方案，針對分布式集群或者數據量級較大時會顯得心有余而力不足，而ELK的出現可以使我們從容面對新的挑戰。

• Logstash：負責日志的收集，處理和儲存
• Elasticsearch：負責日志檢索和分析
• Kibana：負責日志的可視化

ELK(Elasticsearch + Logstash + Kibana)

更新記錄

2019年07月02日 - 轉載同事整理的ELK Stack進行重構
2015年08月31日 - 初稿

閱讀原文 - https://wsgzao.github.io/post/elk/

擴展閱讀

elastic - https://www.elastic.co/cn/
ELK - https://fainyang.github.io/post/elk/

ELK簡介

ELK 官方文檔 是一個分布式、可擴展、實時的搜索與數據分析引擎。目前我在工作中只用來收集 server 的 log, 開發鍋鍋們 debug 的好助手。

安裝設置單節點 ELK

如果你想快速的搭建單節點 ELK, 那么使用 docker 方式肯定是你的最佳選擇。使用三合一的鏡像，文檔詳情
注意：安裝完 docker, 記得設置 mmap counts 大小至少 262144
什么是 mmap

# 設置 mmap 命令
# 臨時添加法
sysctl -w vm.max_map_count=262144  

# 寫入 sysctl.conf 文件里
vim /etc/sysctl.conf

vm.max_map_count=262144  
# 保存好文件執行以下命令

sysctl -p

# 安裝 docker
sudo yum install -y yum-utils device-mapper-persistent-data lvm2

sudo yum-config-manager --add-repo https://download.docker.com/linux/centos/docker-ce.repo

sudo yum install -y docker-ce

sudo systemctl start docker

單節點的機器，不必暴露 9200(Elasticsearch JSON interface) 和 9300(Elasticsearch transport interface) 端口。
如果想在 docker 上暴露端口，用 -p 如果沒有填寫監聽的地址，默認是 0.0.0.0 所有的網卡。建議還是寫明確監聽的地址，安全性更好。

-p 監聽的IP:宿主機端口:容器內的端口
-p 192.168.10.10:9300:9300

命令行啟動一個 ELK

sudo docker run -p 5601:5601 -p 5044:5044 \
-v /data/elk-data:/var/lib/elasticsearch  \
-v /data/elk/logstash:/etc/logstash/conf.d  \
-it -e TZ="Asia/Singapore" -e ES_HEAP_SIZE="20g"  \
-e LS_HEAP_SIZE="10g" --name elk-ubuntu sebp/elk

將配置和數據掛載出來，即使 docker container 出現了問題?？梢粤⒓翠N毀再重啟一個，服務受影響的時間很短。

# 注意掛載出來的文件夾的權限問題
chmod 755 /data/elk-data 
chmod 755 /data/elk/logstash
chown -R root:root /data 
-v /data/elk-data:/var/lib/elasticsearch   # 將 elasticsearch 存儲的數據掛載出來，數據持久化。
-v /data/elk/logstash:/etc/logstash/conf.d # 將 logstash 的配置文件掛載出來，方便在宿主機上修改。

elasticsearch 重要的參數調優

1. ES_HEAP_SIZE Elasticsearch will assign the entire heap specified in jvm.options via the Xms (minimum heap size) and Xmx (maximum heap size) settings. You should set these two settings to be equal to each other. Set Xmx and Xms to no more than 50% of your physical RAM.the exact threshold varies but is near 32 GB. the exact threshold varies but 26 GB is safe on most systems, but can be as large as 30 GB on some systems.
   利弊關系: The more heap available to Elasticsearch, the more memory it can use for its internal caches, but the less memory it leaves available for the operating system to use for the filesystem cache. Also, larger heaps can cause longer garbage collection pauses.
2. LS_HEAP_SIZE 如果 heap size 過低，會導致 CPU 利用率到達瓶頸，造成 JVM 不斷的回收垃圾。 不能設置 heap size 超過物理內存。 至少留 1G 給操作系統和其他的進程。

只需要配置logstash

接下來，我們再來看一看 logstash.conf 記得看注釋
參考鏈接:

1. SSL詳情可參考
2. grok 正則捕獲
3. grok插件語法介紹
4. logstash 配置語法
5. grok 內置 pattern
6. Logstash詳細記錄

input {
  beats {
    port => 5044
    #ssl => true
    #ssl_certificate => "/etc/logstash/logstash.crt"
    #ssl_key => "/etc/logstash/logstash.key"
# 1. SSL詳情可參考 
  }
}
# filter 模塊主要是數據預處理，提取一些信息，方便 elasticsearch 好歸類存儲。
# 2. grok 正則捕獲
# 3. grok插件語法介紹 
# 4. logstash 配置語法 
# 5. grok 內置 pattern 
filter {
    grok {  
      match => {"message" => "%{EXIM_DATE:timestamp}\|%{LOGLEVEL:log_level}\|%{INT:pid}\|%{GREEDYDATA}"}
# message 字段是 log 的內容，例如 2018-12-11 23:46:47.051|DEBUG|3491|helper.py:85|helper._save_to_cache|shop_session
# 在這里我們提取出了 timestamp log_level pid，grok 有內置定義好的patterns: EXIM_DATE, EXIM_DATE, INT
# GREEDYDATA 貪婪數據，代表任意字符都可以匹配 
    }
# 我們在 filebeat 里面添加了這個字段[fields][function]的話，那就會執行對應的 match 規則去匹配 path
# source 字段就是 log 的來源路徑，例如 /var/log/nginx/feiyang233.club.access.log
# match 后我們就可以得到 path=feiyang233.club.access
    if [fields][function]=="nginx" {
        grok {         
        match => {"source" => "/var/log/nginx/%{GREEDYDATA:path}.log%{GREEDYDATA}"}  
            }
        } 
# 例如 ims 日志來源是 /var/log/ims_logic/debug.log
# match 后我們就可以得到 path=ims_logic
    else if [fields][function]=="ims" {
        grok {
        match => {"source" => "/var/log/%{GREEDYDATA:path}/%{GREEDYDATA}"}
            }
        }  

    else {
        grok {
        match => {"source" => "/var/log/app/%{GREEDYDATA:path}/%{GREEDYDATA}"}
            }         
        }
# filebeat 有定義 [fields][function] 時，我們就添加上這個字段，例如 QA
    if [fields][function] {
          mutate {
              add_field => {
                  "function" => "%{[fields][function]}"
                }
            }
        } 
# 因為線上的機器更多，線上的我默認不在 filebeat 添加 function，所以 else 我就添加上 live  
    else {
          mutate {
              add_field => {
                  "function" => "live"
                }
            }
        }
# 在之前 filter message 時，我們得到了 timestamp，這里我們修改一下格式，添加上時區。
    date {
      match => ["timestamp" , "yyyy-MM-dd HH:mm:ss Z"]
      target => "@timestamp"
      timezone => "Asia/Singapore"
    }
# 將之前獲得的 path 替換其中的 / 替換為 - , 因為 elasticsearch index name 有要求
# 例如 feiyang/test  feiyang_test 
    mutate {
     gsub => ["path","/","-"]
      add_field => {"host_ip" => "%{[fields][host]}"}
      remove_field => ["tags","@version","offset","beat","fields","exim_year","exim_month","exim_day","exim_time","timestamp"]
    }
# remove_field 去掉一些多余的字段
}
# 單節點 output 就在本機，也不需要 SSL, 但 index 的命名規則還是需要非常的注意
output {
  elasticsearch {
    hosts => ["localhost:9200"]
    index => "sg-%{function}-%{path}-%{+xxxx.ww}"
# sg-nginx-feiyang233.club.access-2019.13  ww代表周數
  }
}

最終的流程圖如下所示

image

• Lowercase only
• Cannot include , /, *, ?, ", <, >, |, ` ` (space character), ,, #
• Indices prior to 7.0 could contain a colon (:), but that’s been deprecated and won’t be supported in 7.0+
• Cannot start with -, _, +
• Cannot be . or ..
• Cannot be longer than 255 bytes (note it is bytes, so multi-byte characters will count towards the 255 limit faster)

filebeat 配置

在 client 端，我們需要安裝并且配置 filebeat 請參考
Filebeat 模塊與配置
配置文件 filebeat.yml

filebeat.inputs:
- type: log
  enabled: true
  paths: # 需要收集的日志
    - /var/log/app/**  ## ** need high versiob filebeat can support recursive

  fields: #需要添加的字段
    host: "{{inventory_hostname}}" 
    function: "xxx"
  multiline:  # 多行匹配
    match: after
    negate: true  # pay attention the format
    pattern: '^\[[0-9]{4}-[0-9]{2}-[0-9]{2}'   #\[
  ignore_older: 24h
  clean_inactive: 72h

output.logstash:
  hosts: ["{{elk_server}}:25044"]
  # ssl:
  #   certificate_authorities: ["/etc/filebeat/logstash.crt"]

批量部署 filebeat.yml 最好使用 ansible

---
- hosts: all
  become: yes
  gather_facts: yes
  tasks:
  - name: stop filebeat
    service: 
      name: filebeat
      state: stopped
      enabled: yes
      
  - name: upload filebeat.yml 
    template:
     src: filebeat.yml
     dest: /etc/filebeat/filebeat.yml
     owner: root
     group: root
     mode: 0644      

  - name: remove
    file: #delete all files in this directory
      path: /var/lib/filebeat/registry    
      state: absent

  - name: restart filebeat
    service: 
      name: filebeat
      state: restarted
      enabled: yes

查看 filebeat output

首先需要修改配置，將 filebeat 輸出到本地的文件，輸出的格式為 json.

filebeat.inputs:
- type: log
  enabled: true
  paths:
     - /var/log/app/**
  fields:
    host: "x.x.x.x"
    region: "sg"
  multiline:
    match: after
    negate: true
    pattern: '^[0-9]{4}-[0-9]{2}-[0-9]{2}'
  ignore_older: 24h
  clean_inactive: 72h

output.file:
 path: "/home/feiyang"
  filename: feiyang.json

通過上述的配置，我們就可以在路徑 /home/feiyang 下得到輸出結果文件 feiyang.json 在這里需要注意的是，不同版本的 filebeat 輸出結果的格式會有所不同，這會給 logstash 解析過濾造成一點點困難。下面舉例說明 6.x 和 7.x filebeat 輸出結果的不同

{
  "@timestamp": "2019-06-27T15:53:27.682Z",
  "@metadata": {
    "beat": "filebeat",
    "type": "doc",
    "version": "6.4.2"
  },
  "fields": {
    "host": "x.x.x.x",
    "region": "sg"
  },
  "host": {
    "name": "x.x.x.x"
  },
  "beat": {
    "name": "x.x.x.x",
    "hostname": "feiyang-localhost",
    "version": "6.4.2"
  },
  "offset": 1567983499,
  "message": "[2019-06-27T22:53:25.756327232][Info][@http.go.177] [48552188]request",
  "source": "/var/log/feiyang/scripts/all.log"
}

6.4 與 7.2 還是有很大的差異，在結構上。

{
  "@timestamp": "2019-06-27T15:41:42.991Z",
  "@metadata": {
    "beat": "filebeat",
    "type": "_doc",
    "version": "7.2.0"
  },
  "agent": {
    "id": "3a38567b-e6c3-4b5a-a420-f0dee3a3bec8",
    "version": "7.2.0",
    "type": "filebeat",
    "ephemeral_id": "b7e3c0b7-b460-4e43-a9af-6d36c25eece7",
    "hostname": "feiyang-localhost"
  },
  "log": {
    "offset": 69132192,
    "file": {
      "path": "/var/log/app/feiyang/scripts/info.log"
    }
  },
  "message": "2019-06-27 22:41:25.312|WARNING|14186|Option|data|unrecognized|fields=set([u'id'])",
  "input": {
    "type": "log"
  },
  "fields": {
    "region": "sg",
    "host": "x.x.x.x"
  },
  "ecs": {
    "version": "1.0.0"
  },
  "host": {
    "name": "feiyang-localhost"
  }
}

Kibana 簡單的使用

在搭建 ELK 時，暴露出來的 5601 端口就是 Kibana 的服務。
訪問 http://your_elk_ip:5601

image

安裝設置集群 ELK 版本 6.7

ELK 安裝文檔集群主要是高可用，多節點的 Elasticsearch 還可以擴容。本文中用的官方鏡像 The base image is centos:7

Elasticsearch 多節點搭建

官方安裝文檔 Elasticsearch

# 掛載出來的文件夾權限非常的重要
mkdir -p /data/elk-data && chmod 755 /data/elk-data
chown -R root:root /data 
docker run -p WAN_IP:9200:9200 -p 10.66.236.116:9300:9300 \
-v /data/elk-data:/usr/share/elasticsearch/data \
--name feiy_elk \
docker.elastic.co/elasticsearch/elasticsearch:6.7.0

接下來是修改配置文件 elasticsearch.yml

# Master 節點 node-1
# 進入容器 docker exec -it [container_id] bash
# docker exec -it 70ada825aae1 bash
# vi /usr/share/elasticsearch/config/elasticsearch.yml
cluster.name: "feiy_elk"
network.host: 0.0.0.0
node.master: true
node.data: true
node.name: node-1
network.publish_host: 10.66.236.116
discovery.zen.ping.unicast.hosts: ["10.66.236.116:9300","10.66.236.118:9300","10.66.236.115:9300"]

# exit
# docker restart  70ada825aae1

# slave 節點 node-2
# 進入容器 docker exec -it [container_id] bash
# vi /usr/share/elasticsearch/config/elasticsearch.yml
cluster.name: "feiy_elk"
network.host: "0.0.0.0"
node.name: node-2
node.data: true
network.publish_host: 10.66.236.118
discovery.zen.ping.unicast.hosts: ["10.66.236.116:9300","10.66.236.118:9300","10.66.236.115:9300"]

# exit
# docker restart  70ada825aae1

# slave 節點 node-3
# 進入容器 docker exec -it [container_id] bash
# vi /usr/share/elasticsearch/config/elasticsearch.yml
cluster.name: "feiy_elk"
network.host: "0.0.0.0"
node.name: node-3
node.data: true
network.publish_host: 10.66.236.115
discovery.zen.ping.unicast.hosts: ["10.66.236.116:9300","10.66.236.118:9300","10.66.236.115:9300"]

# exit
# docker restart  70ada825aae1

檢查集群節點個數，狀態等

# curl http://wan_ip:9200/_cluster/health?pretty
{
  "cluster_name" : "feiy_elk",
  "status" : "green",
  "timed_out" : false,
  "number_of_nodes" : 3,
  "number_of_data_nodes" : 3,
  "active_primary_shards" : 9,
  "active_shards" : 18,
  "relocating_shards" : 0,
  "initializing_shards" : 0,
  "unassigned_shards" : 0,
  "delayed_unassigned_shards" : 0,
  "number_of_pending_tasks" : 0,
  "number_of_in_flight_fetch" : 0,
  "task_max_waiting_in_queue_millis" : 0,
  "active_shards_percent_as_number" : 100.0
}

最終結果圖在 kibana 上可以看到集群狀態
[圖片上傳失敗...(image-812ccc-1562318412515)]

Kibana 搭建

官方安裝文檔 Kibana

# docker run --link YOUR_ELASTICSEARCH_CONTAINER_NAME_OR_ID:elasticsearch -p 5601:5601 {docker-repo}:{version}
docker run -p 外網IP:5601:5601 --link elasticsearch容器的ID:elasticsearch docker.elastic.co/kibana/kibana:6.7.0

# 注意的是 --link 官方其實并不推薦的，推薦的是 use user-defined networks https://docs.docker.com/network/links/
# 測試不用 --link 也可以通。直接用容器的 IP
docker run -p 外網IP:5601:5601  docker.elastic.co/kibana/kibana:6.7.0

we recommend that you use user-defined networks to facilitate communication between two containers instead of using --link

# vi /usr/share/kibana/config/kibana.yml
# 需要把 hosts IP 改為 elasticsearch 容器的 IP
# 我這里 elasticsearch 容器的 IP 是 172.17.0.2
# 如何查看 docker inspect elasticsearch_ID
server.name: kibana
server.host: "0.0.0.0"
elasticsearch.hosts: [ "http://172.17.0.2:9200" ]
xpack.monitoring.ui.container.elasticsearch.enabled: true

# 退出容器并重啟
docker restart [container_ID]

Logstash 搭建

官方安裝文檔 Logstash

# docker -d 以后臺的方式啟動容器  --name 參數顯式地為容器命名
docker run -p 5044:5044 -d --name test_logstash  docker.elastic.co/logstash/logstash:6.7.0
# 也可以指定網卡，監聽在內網或者外網 監聽在內網 192.168.1.2
docker run -p 192.168.1.2:5044:5044 -d --name test_logstash  docker.elastic.co/logstash/logstash:6.7.0

# vi /usr/share/logstash/pipeline/logstash.conf
# 配置詳情請參考下面的鏈接,記得 output hosts IP 指向 Elasticsearch 的 IP
# Elasticsearch 的默認端口是9200，在下面的配置中可以省略。
hosts => ["IP Address 1:port1", "IP Address 2:port2", "IP Address 3"]

logstash 過濾規則 見上文的配置和 grok 語法規則

# vi /usr/share/logstash/config/logstash.yml
# 需要把 url 改為 elasticsearch master 節點的 IP
http.host: "0.0.0.0"
xpack.monitoring.elasticsearch.url: http://elasticsearch_master_IP:9200
node.name: "feiy"
pipeline.workers: 24 # same with cores

改完配置 exit 從容器里退出到宿主機，然后重啟這個容器。更多配置詳情，參見官方文檔

# 如何查看 container_ID
docker ps -a

docker restart [container_ID]

容災測試

我們把當前的 master 節點 node-1 關機，通過 kibana 看看集群的狀態是怎樣變化的。

image

image

kibana 控制臺 Console

Quick intro to the UI
The Console UI is split into two panes: an editor pane (left) and a response pane (right). Use the editor to type requests and submit them to Elasticsearch. The results will be displayed in the response pane on the right side.

Console understands requests in a compact format, similar to cURL:

# index a doc
PUT index/type/1
{
  "body": "here"
}

# and get it ...
GET index/type/1

While typing a request, Console will make suggestions which you can then accept by hitting Enter/Tab. These suggestions are made based on the request structure as well as your indices and types.

A few quick tips, while I have your attention

• Submit requests to ES using the green triangle button.
• Use the wrench menu for other useful things.
• You can paste requests in cURL format and they will be translated to the Console syntax.
• You can resize the editor and output panes by dragging the separator between them.
• Study the keyboard shortcuts under the Help button. Good stuff in there!

Console 常用的命令

Kibana 控制臺
ELK技術棧中的那些查詢語法

GET _search
{
  "query": {
    "match_all": {}
  }
}

GET /_cat/health?v

GET /_cat/nodes?v

GET /_cluster/allocation/explain

GET /_cluster/state

GET /_cat/thread_pool?v

GET /_cat/indices?health=red&v

GET /_cat/indices?v

#將當前所有的 index 的 replicas 設置為 0

PUT /*/_settings
{
   "index" : {
       "number_of_replicas" : 0,
       "refresh_interval": "30s"
   }
}

GET /_template


# 在單節點的時候，不需要備份，所以將 replicas 設置為 0
PUT _template/app-logstash
{
 "index_patterns": ["app-*"],
 "settings": {
   "number_of_shards": 3,
   "number_of_replicas": 0,
   "refresh_interval": "30s"
  }
}

Elasticsearch 數據遷移

Elasticsearch 數據遷移官方文檔感覺不是很詳細。容器化的數據遷移，我太菜用 reindex 失敗了，snapshot 也涼涼。
最后是用一個開源工具 An Elasticsearch Migration Tool 進行數據遷移的。

wget https://github.com/medcl/esm-abandoned/releases/download/v0.4.2/linux64.tar.gz
tar -xzvf linux64.tar.gz
./esm  -s http://127.0.0.1:9200   -d http://192.168.21.55:9200 -x index_name  -w=5 -b=10 -c 10000 --copy_settings --copy_mappings --force  --refresh

Nginx 代理轉發

因為有時候 docker 重啟，iptables restart 也會刷新，所以導致了我們的限制規則會被更改，出現安全問題。這是由于 docker 的網絡隔離基于 iptable 實現造成的問題。為了避免這個安全問題，我們可以在啟動 docker 時，就只監聽在內網，或者本地 127.0.0.1 然后通過 nginx 轉發。

# cat kibana.conf
server {

    listen 25601;
    server_name x.x.x.x;
    access_log /var/log/nginx/kibana.access.log;
    error_log /var/log/nginx/kibana.error.log;

    location / {
        allow x.x.x.x;
        allow x.x.x.x;
        deny all;

        proxy_http_version 1.1;
        proxy_buffer_size 64k;
        proxy_buffers   32 32k;
        proxy_busy_buffers_size 128k;
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP       $remote_addr;
        proxy_set_header X-Forwarded-For  $proxy_add_x_forwarded_for;

        proxy_pass    http://127.0.0.1:5601;

    }
}

! 這里需要注意的是， iptable filter 表 INPUT 鏈 有沒有阻擋 172.17.0.0/16 docker 默認的網段。是否阻擋了 25601 這個端口。

踩過的坑

• iptables 防不住。需要看上一篇博客里的 iptable 問題?；蛘弑O聽在內網，用 Nginx 代理轉發。
• elk 網絡問題
• elk node
• discovery.type=single-node 在測試單點時可用，搭建集群時不能設置這個環境變量，詳情見官方文檔
• ELK的一次吞吐量優化
• filebeat 版本過低導致 recursive glob patterns ** 不可用
  用 ansible 升級 filebeat

---
- hosts: all
  become: yes
  gather_facts: yes
  tasks:
  - name: upload filebeat.repo 
    copy:
     src: elasticsearch.repo
     dest: /etc/yum.repos.d/elasticsearch.repo
     owner: root
     group: root
     mode: 0644

  - name: install the latest version of filebeat
    yum:
      name: filebeat
      state: latest

  - name: restart filebeat
    service: 
      name: filebeat
      state: restarted
      enabled: yes
      
# elasticsearch.repo
[elasticsearch-6.x]
name=Elasticsearch repository for 6.x packages
baseurl=https://artifacts.elastic.co/packages/6.x/yum
gpgcheck=1
gpgkey=https://artifacts.elastic.co/GPG-KEY-elasticsearch
enabled=1
autorefresh=1
type=rpm-md

• filebeat 7.x 與 6.x 不兼容問題. 關鍵字變化很大, 比如說 "sorce" 變為了 [log][file][path]

參考文章

1. 騰訊云Elasticsearch Service 這個騰訊云的專欄非常的不錯，請您一定要點開看一眼，總有你想要的。
2. ELK重難點總結和整體優化配置