ESRally性能测试步骤

Rocky大约 7 分钟

ES7.4.1搭建

ES统一采用7.4.1版本

ES的Docker镜像部署

Docker的安装这里不再赘述

单节点部署

  • vi docker-compose-es-single-node.yml

    version: "2"
    services:
      es-single-node:
    	image: elasticsearch:7.4.1
    	container_name: es-single-node
    	environment:
    	  - node.name=es-single-node
    	  - cluster.name=es-cluster
    	  - cluster.initial_master_nodes=es-single-node
    	  - discovery.seed_hosts=es-single-node
    	  - discovery.zen.minimum_master_nodes=1
    	  - node.master=true
    	  - node.data=true
    	  - http.port=9200
    	  - transport.tcp.port=9300
    	  - http.cors.enabled=true
    	  - http.cors.allow-origin="*"
    	  - "ES_JAVA_OPTS=-Xms128m -Xmx128m"
    	ports:
    	  - 9210:9200
    	  - 9310:9300
    
  • docker-compose -f docker-compose-es-single-node.yml up -d (这个步骤也可以用docker的界面管理工具portainer完成)

多节点部署

这里演示2个master2个data节点的配置,增加删除节点依葫芦画瓢即可

  • vi docker-compose-2master2data.yml

    version: "2"
    services:
      master01:
    	image: elasticsearch:7.4.1
    	container_name: master01
    	environment:
    	  - node.name=master01
    	  - cluster.name=es-cluster
    	  - cluster.initial_master_nodes=master01,master02
    	  - discovery.seed_hosts=master01,master02,data01,data02
    	  - discovery.zen.minimum_master_nodes=2
    	  - node.master=true
    	  - node.data=false
    	  - http.port=9200
    	  - transport.tcp.port=9300
    	  - http.cors.enabled=true
    	  - http.cors.allow-origin="*"
    	  - "ES_JAVA_OPTS=-Xms128m -Xmx128m"
    	ports:
    	  - 9200:9200
    	  - 9300:9300
      master02:
    	image: elasticsearch:7.4.1
    	container_name: master02
    	environment:
    	  - node.name=master02
    	  - cluster.name=es-cluster
    	  - cluster.initial_master_nodes=master01,master02
    	  - discovery.seed_hosts=master01,master02,data01,data02
    	  - discovery.zen.minimum_master_nodes=2
    	  - node.master=true
    	  - node.data=false
    	  - http.port=9200
    	  - transport.tcp.port=9300
    	  - http.cors.enabled=true
    	  - http.cors.allow-origin="*"
    	  - "ES_JAVA_OPTS=-Xms128m -Xmx128m"
    	ulimits:
    	  memlock:
    		soft: -1
    		hard: -1
    	ports:
    	  - 9201:9200
    	  - 9301:9300
      data01:
    	image: elasticsearch:7.4.1
    	container_name: data01
    	environment:
    	  - node.name=data01
    	  - cluster.name=es-cluster
    	  - cluster.initial_master_nodes=master01,master02
    	  - discovery.seed_hosts=master01,master02,data01,data02
    	  - discovery.zen.minimum_master_nodes=2
    	  - node.master=false
    	  - node.data=true
    	  - http.port=9200
    	  - transport.tcp.port=9300
    	  - http.cors.enabled=true
    	  - http.cors.allow-origin="*"
    	  - "ES_JAVA_OPTS=-Xms128m -Xmx128m"
    	ulimits:
    	  memlock:
    		soft: -1
    		hard: -1
    	ports:
    	  - 9202:9200
    	  - 9302:9300
      data02:
    	image: elasticsearch:7.4.1
    	container_name: data02
    	environment:
    	  - node.name=data02
    	  - cluster.name=es-cluster
    	  - cluster.initial_master_nodes=master01,master02
    	  - discovery.seed_hosts=master01,master02,data01,data02
    	  - discovery.zen.minimum_master_nodes=2
    	  - node.master=false
    	  - node.data=true
    	  - http.port=9200
    	  - transport.tcp.port=9300
    	  - http.cors.enabled=true
    	  - http.cors.allow-origin="*"
    	  - "ES_JAVA_OPTS=-Xms128m -Xmx128m"
    	ulimits:
    	  memlock:
    		soft: -1
    		hard: -1
    	ports:
    	  - 9203:9200
    	  - 9303:9300
      kibana:
    	image: kibana:7.4.1
    	environment:
    	  - ELASTICSEARCH_HOSTS=["http://master01:9200","http://master02:9200","http://data01:9200","http://data02:9200"]
    	ports:
    	  - 5601:5601
    
  • docker-compose -f docker-compose-2master2data.yml up -d

Centos部署

单节点部署

  1. 下载es7.4.1安装文件到服务器并解压到 /usr/share/elasticsearch 目录,然后已root用户执行:
    echo "* soft nofile 125536" >> /etc/security/limits.conf
    echo "* hard nofile 125536" >> /etc/security/limits.conf
    echo "* soft nproc 8096" >> /etc/security/limits.conf
    echo "* hard nproc 8096" >> /etc/security/limits.conf
    echo "* soft memlock unlimited" >> /etc/security/limits.conf
    echo "* hard memlock unlimited" >> /etc/security/limits.conf
    
    echo "vm.max_map_count=522144" >> /etc/sysctl.conf
    sysctl -p
    
    adduser elastic
    chown -R elastic:elastic /usr/share/elasticsearch
    
  2. 修改elasticsearch.yml,可参考下面的配置
    # 这两个配置都写master节点的ip
    cluster.initial_master_nodes: ["192.168.1.55"]
    discovery.seed_hosts: ["192.168.1.55"]
    
    # 节点名称,其余两个节点分别为node-ip
    # 节点名称以node-开头,以当前节点IP结尾
    node.name: master192.168.1.55
    
    discovery.zen.minimum_master_nodes: 1
    
    # 指定该节点是否有资格被选举成为master节点,默认是true,es是默认集群中的第一台机器为master,如果这台机挂了就会重新选举master
    node.master: true
    
    # 允许该节点存储数据(默认开启)
    node.data: true
    
    # 绑定的ip地址
    network.host: 0.0.0.0
    
    # 集群的名称
    cluster.name: xxxxx-dssa
    
    # 索引数据的存储路径
    path.data: data
    
    # 日志文件的存储路径
    path.logs: logs
    
    # 快照仓库路径
    path.repo: ["/opt/xxxxx/backups/es_repo","/opt/xxxxx/backups/rsyslog-audit","/opt/xxxxx/backups/rsyslog-firewall","/opt/xxxxx/backups/rsyslog-desens","/opt/xxxxx/backups/rsyslog-encrypt"]
    
    # 正式部署需要设置为true来锁住内存。因为内存交换到磁盘对服务器性能来说是致命的,当jvm开始swapping时es的效率会降低,所以要保证它不swap
    bootstrap.memory_lock: true
    
    # 设置对外服务的http端口,默认为9200
    http.port: 9200
    
    # 设置节点间交互的tcp端口,默认是9300
    transport.tcp.port: 9300
    
    # 如果没有足够大的内存,因为了elasticsearch引用文件,系统内存会大量用于系统cache(linux的内存管理机制)。
    # 由于系统cache释放缓慢,而导致这个过程非常长,这有可能使你的节点GC非常频繁,从而导致集群不稳定。
    # 建议把bootstrap.mlockall设为true
    # 这个参数在7.4.1无效
    # bootstrap.mlockall: true
    
    # 开启跨域访问
    http.cors.enabled: true
    http.cors.allow-origin: "*"
    
    
  3. 修改jvm.options,一般把Xms Xmx内存配置为可用内存的一半
  4. 尝试启动观察日志看看有没有问题
    su elastic
    cd /usr/share/elasticsearch
    bin/elasticsearch    前台执行
    bin/elasticsearch    后台执行,需要到配置的日志目录去看日志
    
  5. 开机启动
    • vi /usr/lib/systemd/system/elasticsearch.service 输入如下内容:
    [Unit]
    Description=elasticsearch service
    
    [Service]
    User=elastic
    ExecStart=/usr/share/elasticsearch/bin/elasticsearch
    
    [Install]
    WantedBy=multi-user.target
    
    • systemctl enable elasticsearch (开机启动)
    • systemctl restart elasticsearch (重启es)

多节点部署

多节点部署和单节点部署几乎一模一样,只需要根据自身要搭建的集群节点规划修改下配置文件即可

一般修改如下几个配置:

cluster.initial_master_nodes: ["master01","master02"]
discovery.seed_hosts: ["master01","master02","data01","data02"]

node.name: 节点的名字

# 计算公式是这样的: master候选节点的个数/2 + 1,设置不恰当可能存在脑裂问题
discovery.zen.minimum_master_nodes: 2

# 指定该节点是否有资格被选举成为master节点,默认是true,es是默认集群中的第一台机器为master,如果这台机挂了就会重新选举master
node.master: true

# 允许该节点存储数据(默认开启)
node.data: true

ESRally

esrally是es官方提供的对es进行性能测试的工具 官网:https://esrally.readthedocs.io/en/stable/quickstart.htmlopen in new window

esrally安装

直接拉取docker镜像即可:

1. 修改host文件,添加如下配置
172.16.1.201 prod.docker
172.16.1.201 dev.docker

2. 修改docker配置文件,增加如下配置:
 { "insecure-registries": ["prod.docker:8085", "dev.docker:8085"] }

3. 拉取docker esrally镜像
docker pull dev.docker:8085/elastic/rally:2.0.0

4. 执行测试
docker run --rm -i -v /opt/esrally:/opt/esrally dev.docker:8085/elastic/rally:2.0.0  --pipeline=benchmark-only --track-path=/opt/esrally --target-hosts=192.168.1.55:9200

enter description here
enter description here

准备es的索引模板

在 tfs 中已经准备了几个我们生产环境使用的es索引模板,可根据自身情况自行修改

enter description here
enter description here

由于索引模板是yml格式的,需要转换为json格式:通过修改index.sh中的相关参数并执行index.sh即可完成json转换。

index.sh完成了两个动作(可根据自身情况自行删除): - yml转换为json - 连接es创建索引

准备测试数据

在tfs中也已经准备好了生成测试数据的工具

enter description here
enter description here

只需要修改eslog_generator.sh中的三个参数即可根据索引模板生产测试数据

编辑track.json文件

这个文件是esrally需要使用的,可参考:

{
    "version": 2,
    "description": "xxxxx logs",
    "indices": [
        {
            "name": "xxxxx_audit_bigdata"
        }
    ],
    "corpora": [
        {
            "name": "http_logs",
            "documents": [
                {
                    "source-file": "xxxxx_audit_bigdata1200w.json.bz2",
                    "document-count": 12000000,
					"uncompressed-bytes": 30760
                }
            ]
        }
    ],
    "challenges": [
        {
            "name": "index-test",
            "default": true,
            "schedule": [
            	
              	{
                    "operation": {
                        "name": "putdata_bulk",
                        "operation-type": "bulk",
                        "bulk-size": 1000
                    },
                    "clients": 5
                },
                {
                    "operation": {
                        "name": "search",
                        "operation-type": "search",
                        "body": {
                            "query":{
                                "match_all": {}
                            }
                        }
                    },
                    "clients": 5
                }
            ]
        }
    ]
}

enter description here
enter description here

执行

track.json文件和生成的数据文件要放在同一个目录,假设为 /opt/esrally/track,则执行

esrally race --pipeline=benchmark-only --track-path=/opt/esrally/track --target-hosts=esip:9200

执行完成会有下面的一个结果输出

enter description here
enter description here

系统推荐









  • 随机毒鸡汤:天下没有免费的午餐,但是有免费的,苦可以吃啊。