mongodb的分片功能是建立在副本集之上的,所以首先我们尝试着配置副本集。
docker启动3个已经安装好mongo的镜像
# docker run -idt --name mongodb_01 mongodb_master:v2 /bin/bash # docker run -idt --name mongodb_02 mongodb_master:v2 /bin/bash # docker run -idt --name mongodb_03 mongodb_master:v2 /bin/bash
查看容器ip
# docker inspect mongodb_01 | grep IP
3个容器的ip为
172.17.0.4,172.17.0.5,172.17.0.6
进入容器,分别建立mongodb的数据和日志文件夹,编辑配置文件
# docker exec -it mongodb_01 /bin/bash # mkdir -p /opt/mongodb/rs0/data # mkdir -p /opt/mongodb/rs0/log
# vi /usr/local/mongdb/conf/rs0.conf
dbpath=/opt/mongodb/rs0/data #数据目录
logpath=/opt/mongodb/rs0/log/rs0.log #日志目录
pidfilepath=/opt/mongodb/rs0/log/rs0.pid #pid
logappend=true
replSet=rs0 #副本集名称
bind_ip=172.17.0.4 #容器对应的ip
port=27617
fork=true
maxConns=2000
//启动容器
# mongod -f /usr/local/mongdb/conf/rs0.conf
3个容器的mongodb全部启动后,随便连接一个mongodb
# mongo --host 172.17.0.4 --port 27617 > rs.initiate() //初始化副本集 > rs.conf()//确认更改 > rs.add({host:"172.17.0.5:27518", priority: 6}) //将另外两个mongo服务加入副本集 > rs.conf() //确认更改 > rs.status() //查看副本及状态
priority代表副本集的优先级,数值越大优先级越高,可以通过rs.status()查看当前副本集的状态,stateStr表示副本及的身份,可以看到172.17.0.5目前的身份是PRIMARY
,另外两个都是SECONDARY,这时我们停止172.17.0.5的mongo服务,过一段时间再看,0.4和0.6中其中一台节点就会变成PRIMARY
,再开启0.5上的mongo服务,又会回复成原来的状态
"members" : [ { "_id" : 0, "name" : "172.17.0.6:27619", "health" : 1, "state" : 2, "stateStr" : "SECONDARY", "uptime" : 264637, "optime" : { "ts" : Timestamp(1539406655, 1), "t" : NumberLong(4) }, "optimeDate" : ISODate("2018-10-13T04:57:35Z"), "syncingTo" : "172.17.0.4:27617", "syncSourceHost" : "172.17.0.4:27617", "syncSourceId" : 2, "infoMessage" : "", "configVersion" : 3, "self" : true, "lastHeartbeatMessage" : "" }, { "_id" : 1, "name" : "172.17.0.5:27618", "health" : 1, "state" : 1, "stateStr" : "PRIMARY", "uptime" : 263943, "optime" : { "ts" : Timestamp(1539406655, 1), "t" : NumberLong(4) }, "optimeDurable" : { "ts" : Timestamp(1539406655, 1), "t" : NumberLong(4) }, "optimeDate" : ISODate("2018-10-13T04:57:35Z"), "optimeDurableDate" : ISODate("2018-10-13T04:57:35Z"), "lastHeartbeat" : ISODate("2018-10-13T04:57:38.894Z"), "lastHeartbeatRecv" : ISODate("2018-10-13T04:57:38.892Z"), "pingMs" : NumberLong(0), "lastHeartbeatMessage" : "", "syncingTo" : "", "syncSourceHost" : "", "syncSourceId" : -1, "infoMessage" : "", "electionTime" : Timestamp(1539142727, 1), "electionDate" : ISODate("2018-10-10T03:38:47Z"), "configVersion" : 3 }, { "_id" : 2, "name" : "172.17.0.4:27617", "health" : 1, "state" : 2, "stateStr" : "SECONDARY", "uptime" : 264390, "optime" : { "ts" : Timestamp(1539406655, 1), "t" : NumberLong(4) }, "optimeDurable" : { "ts" : Timestamp(1539406655, 1), "t" : NumberLong(4) }, "optimeDate" : ISODate("2018-10-13T04:57:35Z"), "optimeDurableDate" : ISODate("2018-10-13T04:57:35Z"), "lastHeartbeat" : ISODate("2018-10-13T04:57:38.893Z"), "lastHeartbeatRecv" : ISODate("2018-10-13T04:57:38.893Z"), "pingMs" : NumberLong(0), "lastHeartbeatMessage" : "", "syncingTo" : "172.17.0.5:27618", "syncSourceHost" : "172.17.0.5:27618", "syncSourceId" : 1, "infoMessage" : "", "configVersion" : 3 } ],
了解了副本集的配置,接下来进行分片的配置
分片即是通过某种算法,将数据分散到不同的片区上,但这样同时会产生一个问题,如果某一个片区出现问题则整个数据都会变的不可用。
所以,需要将分散到不同片区上的数据再以副本集的形式存放,这样在分片的同时也就具备了容错的能力,概念上其实和RAID很相似。
mongodb的分片配置包含以下几个角色:
Config Server:配置分片信息
Shard:实际存储分片数据的地方
mongos:分片的路由,也是前端实际连接的实例
这里盗一张网上的图:
依然用之前的3个镜像,建立需要的文件夹
conf对应Config Server的文件夹
mongos 对应 mongos的文件夹
shard这里我们分为3片,分别对应 shard1,shard2,shard3
# mkdir -p /opt/mongodb/conf/data # mkdir -p /opt/mongodb/conf/log # mkdir -p /opt/mongodb/mongos/data # mkdir -p /opt/mongodb/mongos/log # mkdir -p /opt/mongodb/shard1/data # mkdir -p /opt/mongodb/shard1/log # mkdir -p /opt/mongodb/shard2/data # mkdir -p /opt/mongodb/shard2/log # mkdir -p /opt/mongodb/shard3/data # mkdir -p /opt/mongodb/shard3/log
编辑每个角色的配置文件
Config Server
# vi /usr/local/mongdb/conf/conf.conf dbpath=/opt/mongodb/conf/data logpath=/opt/mongodb/conf/log/conf.log pidfilepath=/opt/mongodb/conf/log/conf.pid logappend=true replSet=configs bind_ip=172.17.0.6 port=27019 fork=true maxConns=2000 configsvr=true #Config Server服务器增加此行
mongos
# vi /usr/local/mongdb/conf/mongos.conf logpath=/opt/mongodb/mongos/log/mongos.log pidfilepath=/opt/mongodb/conf/log/mongos.pid logappend=true bind_ip=172.17.0.6 port=27419 fork=true maxConns=2000 configdb=configs/172.17.0.4:27017,172.17.0.5:27018,172.17.0.6:27019 #Config Server的地址
shard
# vi /usr/local/mongdb/conf/shard1.conf pidfilepath = /opt/mongodb/shard1/log/shard1.pid dbpath = /opt/mongodb/shard1/data logpath = /opt/mongodb/shard1/log/shard1.log logappend = true bind_ip = 172.17.0.6 port = 27119 fork = true replSet=shard1 shardsvr = true maxConns=20000
配置好后,分别启动Config Server,shard1,shard2,shard3的 mong实例
按照上面介绍的配置副本集方法,分别给Config server,shard1,shard2,shard3配置副本集
最后启动mongos实例,注意命令是mongos 不是mongod
# mongos -f /usr/local/mongodb/conf/mongos.conf
连接mongos实例启用分片
# mongo --host 172.17.0.6:27419 >sh.addShard("shard1/172.17.0.4:27117,172.17.0.5:27118,172.17.0.6:27119") >sh.addShard("shard1/172.17.0.4:27217,172.17.0.5:27218,172.17.0.6:27219") >sh.addShard("shard1/172.17.0.4:27317,172.17.0.5:27318,172.17.0.6:27319") >db.runCommand( { enablesharding :"testshard"}); //数据库启用分片 >db.runCommand( { shardcollection : "testshard.test",key : {id: "hashed"} } ) //表启用分片,并指定片键
到这里,mongodb的分片集群就配置完成了。可以在不同的镜像中启用mongos和keepalived配合实现高可用。