elasticsearch是一個(gè)基于lucene的搜索服務(wù)器。它提供了一個(gè)分布式多用戶能力的全文搜索引擎，基于restful web接口。elasticsearch是用java開發(fā)的，并作為apache許可條款下的開放源碼發(fā)布，是當(dāng)前流行的企業(yè)級(jí)搜索引擎。
本文將使用docker容器（使用docker-compose編排）快速部署elasticsearch 集群，可用于開發(fā)環(huán)境（單機(jī)多實(shí)例）或生產(chǎn)環(huán)境部署。
注意，6.x版本已經(jīng)不能通過 -epath.config 參數(shù)去指定配置文件的加載位置，文檔說明：
for the archive distributions, the config directory location defaults to $es_home/config. the location of the >config directory can be changed via the es_path_conf environment variable as follows:
es_path_conf=/path/to/my/config ./bin/elasticsearch
alternatively, you can export the es_path_conf environment variable via the command line or via your shell profile.
即交給環(huán)境變量 es_path_conf 來設(shè)定了，單機(jī)部署多個(gè)實(shí)例且不使用容器的同學(xué)多多注意。
準(zhǔn)備工作
安裝 docker & docker-compose
這里推進(jìn)使用 daocloud 做個(gè)加速安裝：
#docker
curl -ssl https://get.daocloud.io/docker | sh
#docker-compose
curl -l \
https://get.daocloud.io/docker/compose/releases/download/1.23.2/docker-compose-`uname -s`-`uname -m` \
> /usr/local/bin/docker-compose
chmod +x /usr/local/bin/docker-compose
#查看安裝結(jié)果
docker-compose -v數(shù)據(jù)目錄
#創(chuàng)建數(shù)據(jù)/日志目錄 這里我們部署3個(gè)節(jié)點(diǎn)
mkdir /opt/elasticsearch/data/{node0,nod1,node2} -p
mkdir /opt/elasticsearch/logs/{node0,nod1,node2} -p
cd /opt/elasticsearch
#權(quán)限我也很懵逼啦 給了 privileged 也不行 索性0777好了
chmod 0777 data/* -r && chmod 0777 logs/* -r
#防止jvm報(bào)錯(cuò)
echo vm.max_map_count=262144 >> /etc/sysctl.conf
sysctl -pdocker-compse 編排服務(wù)
創(chuàng)建編排文件
vim docker-compose.yml
參數(shù)說明
- cluster.name=elasticsearch-cluster
集群名稱
- node.name=node0
- node.master=true
- node.data=true
節(jié)點(diǎn)名稱、是否可作為主節(jié)點(diǎn)、是否存儲(chǔ)數(shù)據(jù)
- bootstrap.memory_lock=true
鎖定進(jìn)程的物理內(nèi)存地址避免交換（swapped）來提高性能
- http.cors.enabled=true
- http.cors.allow-origin=*
開啟cors以便使用head插件
- es_java_opts=-xms512m -xmx512m
jvm內(nèi)存大小配置
- discovery.zen.ping.unicast.hosts=elasticsearch_n0,elasticsearch_n1,elasticsearch_n2
- discovery.zen.minimum_master_nodes=2
由于5.2.1后的版本是不支持多播的，所以需要手動(dòng)指定集群各節(jié)點(diǎn)的tcp數(shù)據(jù)交互地址，用于集群的節(jié)點(diǎn)發(fā)現(xiàn)和failover，默認(rèn)缺省9300端口，如設(shè)定了其它端口需另行指定，這里我們直接借助容器通信，也可以將各節(jié)點(diǎn)的9300映射至宿主機(jī)，通過網(wǎng)絡(luò)端口通信。
設(shè)定failover選取的quorum = nodes/2 + 1
當(dāng)然，也可以掛載自己的配置文件，es鏡像的配置文件是/usr/share/elasticsearch/config/elasticsearch.yml，掛載如下:
volumes:
- path/to/local/elasticsearch.yml:/usr/share/elasticsearch/config/elasticsearch.yml:rodocker-compose.yml
version: '3'
services:
elasticsearch_n0:
image: elasticsearch:6.6.2
container_name: elasticsearch_n0
privileged: true
environment:
- cluster.name=elasticsearch-cluster
- node.name=node0
- node.master=true
- node.data=true
- bootstrap.memory_lock=true
- http.cors.enabled=true
- http.cors.allow-origin=*
- es_java_opts=-xms512m -xmx512m
- discovery.zen.ping.unicast.hosts=elasticsearch_n0,elasticsearch_n1,elasticsearch_n2
- discovery.zen.minimum_master_nodes=2
ulimits:
memlock:
soft: -1
hard: -1
volumes:
- ./data/node0:/usr/share/elasticsearch/data
- ./logs/node0:/usr/share/elasticsearch/logs
ports:
- 9200:9200
elasticsearch_n1:
image: elasticsearch:6.6.2
container_name: elasticsearch_n1
privileged: true
environment:
- cluster.name=elasticsearch-cluster
- node.name=node1
- node.master=true
- node.data=true
- bootstrap.memory_lock=true
- http.cors.enabled=true
- http.cors.allow-origin=*
- es_java_opts=-xms512m -xmx512m
- discovery.zen.ping.unicast.hosts=elasticsearch_n0,elasticsearch_n1,elasticsearch_n2
- discovery.zen.minimum_master_nodes=2
ulimits:
memlock:
soft: -1
hard: -1
volumes:
- ./data/node1:/usr/share/elasticsearch/data
- ./logs/node1:/usr/share/elasticsearch/logs
ports:
- 9201:9200
elasticsearch_n2:
image: elasticsearch:6.6.2
container_name: elasticsearch_n2
privileged: true
environment:
- cluster.name=elasticsearch-cluster
- node.name=node2
- node.master=true
- node.data=true
- bootstrap.memory_lock=true
- http.cors.enabled=true
- http.cors.allow-origin=*
- es_java_opts=-xms512m -xmx512m
- discovery.zen.ping.unicast.hosts=elasticsearch_n0,elasticsearch_n1,elasticsearch_n2
- discovery.zen.minimum_master_nodes=2
ulimits:
memlock:
soft: -1
hard: -1
volumes:
- ./data/node2:/usr/share/elasticsearch/data
- ./logs/node2:/usr/share/elasticsearch/logs
ports:
- 9202:9200這里我們分別為node0/node1/node2開放宿主機(jī)的9200/9201/9202作為http服務(wù)端口，各實(shí)例的tcp數(shù)據(jù)傳輸用默認(rèn)的9300通過容器管理通信。
如果需要多機(jī)部署，則將es的transport.tcp.port: 9300端口映射至宿主機(jī)xxxx端口，discovery.zen.ping.unicast.hosts填寫各主機(jī)代理的地址即可：
#比如其中一臺(tái)宿主機(jī)為192.168.1.100
...
- discovery.zen.ping.unicast.hosts=192.168.1.100:9300,192.168.1.101:9300,192.168.1.102:9300
...
ports:
...
- 9300:9300創(chuàng)建并啟動(dòng)服務(wù)
[root@localhost elasticsearch]# docker-compose up -d
[root@localhost elasticsearch]# docker-compose ps
name command state ports
--------------------------------------------------------------------------------------------
elasticsearch_n0 /usr/local/bin/docker-entr ... up 0.0.0.0:9200->9200/tcp, 9300/tcp
elasticsearch_n1 /usr/local/bin/docker-entr ... up 0.0.0.0:9201->9200/tcp, 9300/tcp
elasticsearch_n2 /usr/local/bin/docker-entr ... up 0.0.0.0:9202->9200/tcp, 9300/tcp
#啟動(dòng)失敗查看錯(cuò)誤
[root@localhost elasticsearch]# docker-compose logs
#最多是一些訪問權(quán)限/jvm vm.max_map_count 的設(shè)置問題查看集群狀態(tài)
192.168.20.6 是我的服務(wù)器地址
訪問http://192.168.20.6:9200/_cat/nodes?v即可查看集群狀態(tài)：
ip heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
172.25.0.3 36 98 79 3.43 0.88 0.54 mdi * node0
172.25.0.2 48 98 79 3.43 0.88 0.54 mdi - node2
172.25.0.4 42 98 51 3.43 0.88 0.54 mdi - node1驗(yàn)證 failover
通過集群接口查看狀態(tài)
模擬主節(jié)點(diǎn)下線，集群開始選舉新的主節(jié)點(diǎn)，并對(duì)數(shù)據(jù)進(jìn)行遷移，重新分片。
[root@localhost elasticsearch]# docker-compose stop elasticsearch_n0
stopping elasticsearch_n0 ... done集群狀態(tài)(注意換個(gè)http端口 原主節(jié)點(diǎn)下線了)，down掉的節(jié)點(diǎn)還在集群中，等待一段時(shí)間仍未恢復(fù)后就會(huì)被剔出
ip heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
172.25.0.2 57 84 5 0.46 0.65 0.50 mdi - node2
172.25.0.4 49 84 5 0.46 0.65 0.50 mdi * node1
172.25.0.3 mdi - node0等待一段時(shí)間
ip heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
172.25.0.2 44 84 1 0.10 0.33 0.40 mdi - node2
172.25.0.4 34 84 1 0.10 0.33 0.40 mdi * node1恢復(fù)節(jié)點(diǎn) node0
[root@localhost elasticsearch]# docker-compose start elasticsearch_n0
starting elasticsearch_n0 ... done等待一段時(shí)間
ip heap.percent ram.percent cpu load_1m load_5m load_15m node.role master name
172.25.0.2 52 98 25 0.67 0.43 0.43 mdi - node2
172.25.0.4 43 98 25 0.67 0.43 0.43 mdi * node1
172.25.0.3 40 98 46 0.67 0.43 0.43 mdi - node0配合 head 插件觀察
git clone git://github.com/mobz/elasticsearch-head.git
cd elasticsearch-head
npm install
npm run start集群狀態(tài)圖示更容易看出數(shù)據(jù)自動(dòng)遷移的過程
1、集群正常 數(shù)據(jù)安全分布在3個(gè)節(jié)點(diǎn)上
2、下線 node1 主節(jié)點(diǎn) 集群開始遷移數(shù)據(jù)
遷移中
遷移完成
3、恢復(fù) node1 節(jié)點(diǎn)
問題小記
elasticsearch watermark
部署完后創(chuàng)建索引發(fā)現(xiàn)有些分片處于 unsigned 狀態(tài)，是由于 elasticsearch watermark：low,high,flood_stage的限定造成的，默認(rèn)硬盤使用率高于85%就會(huì)告警，開發(fā)嘛，手動(dòng)關(guān)掉好了，數(shù)據(jù)會(huì)分片到各節(jié)點(diǎn)，生產(chǎn)自行決斷。
curl -x put http://192.168.20.6:9201/_cluster/settings \
-h 'content-type':'application/json' \
-d '{transient:{cluster.routing.allocation.disk.threshold_enabled: false}}'