一,概述
mongodb是一个典型的文档型nosql数据库,下面说一下Nosql包括的几大类型数据库:
NoSQL的数据存储模型
键值模型:
数据模型:key-value存储
优点:查找速度快
缺点:数据无结构,通常只被当作字符串或二进制数据
应用场景:内容缓存
实例:Redis, Dynamo
列式模型:
数据模型:数据按列存储,将同一列数据存在一起;
优点:查找迅速、可扩展性强、易于实现分布式;
缺点:功能相对SQL很有限;
应用场景:分布式文件系统或分布式存储
实例:Bigtable, Cassandra, HBase, Hypertable
文档模型:
数据模型:与键值模型类似,value指向结构化数据;
优点:数据格式要求不严格,无需事先定义结构
缺点:查询性能不高,缺乏统一查询语法
应用场景:web应用;
实例:MongoDB, CouchDB
图式模型:
数据模型:图结构模型
优点:利用图结构相关算法提高性能,并足特殊场景应用需求
缺点:难以实现分布式,功能有定向性;
应用场景:社交网络、推荐系统、关系图谱
实例:Neo4J
本文介绍的是mongodb的复制集以及sharding,下面来说一下mongodb为什么要做sharding
越来越大的数据集及不断提升吞吐量的应用程序对单台mongodb服务器来讲是一个挑战————大量的查询很快即能耗尽CPU的计算能力,而较大的数据集存储需求也有可能很快超出单节点的存储能力。最终,工作集的大多超出了系统的RAM并给I/O带去巨大压力。数据库管理系统界解决此类问题通常有两类方案:向上扩展和水平扩展。
sharding即是水平扩展的一种解决方案。它通过将数据集分布于多个也称作分片(shard)的节点上来降低单节点的访问压力。每个分片都是一个独立的数据库,所有的分片组合起来构成一个逻辑上的完整意义的数据库。因此,分片机制降低了每个分片的数据操作量及需要存储的数据量。
要构建一个 MongoDB Sharding Cluster,需要三种角色:
Shard Server:用于存储实际的数据块,也就是元数据,实际生产环境中一个shard server角色可由几台机器组个一个relica set承担,防止主机单点故障
Config Server:存储了整个 Cluster Metadata,其中包括 chunk 信息。
Route Server:前端路由,客户端由此接入,且让整个集群看上去像单一数据库,前端应用可以透明使用。
下面是mongodb sharding的工作原理图:
二,实验环境
3台主机的环境:
192.168.30.116 OS:Centos 6.4 x86_64
192.168.30.117 OS:Centos 6.4 x86_64
192.168.30.119 OS:Centos 6.4 x86_64
下面是主机角色:
下面开始安装mongodb,这里使用通用的二进制包安装
[root@node1 ~]# tar zxvf mongodb-linux-x86_64-2.0.4.tgz -C /usr/local/[root@node1 ~]# mv /usr/local/mongodb-linux-x86_64-2.0.4 /usr/local/mongodb[root@node1 ~]# groupadd -r mongod[root@node1 ~]# useradd -M -r -g mongod -d /mongodb/data -s /bin/false -c mongod mongod[root@node1 ~]# mkdir /mongodb/data/shard11 -p[root@node1 ~]# mkdir /mongodb/data/shard21 -p[root@node1 ~]# mkdir /mongodb/data/config -p[root@node1 ~]# mkdir -p /var/log/mongo[root@node1 ~]# chown -R mongod.mongod /mongodb/data/ /var/log/mongo/[root@node2 ~]# tar zxvf mongodb-linux-x86_64-2.0.4.tgz -C /usr/local/[root@node2 ~]# mv /usr/local/mongodb-linux-x86_64-2.0.4 /usr/local/mongodb[root@node2 ~]# groupadd -r mongod[root@node2 ~]# useradd -M -r -g mongod -d /mongodb/data -s /bin/false -c mongod mongod[root@node2 ~]# mkdir /mongodb/data/shard12 -p[root@node2 ~]# mkdir /mongodb/data/shard22 -p[root@node2 ~]# mkdir /mongodb/data/config -p[root@node2 ~]# mkdir -p /var/log/mongo[root@node2 ~]# chown -R mongod.mongod /mongodb/data/ /var/log/mongo/[root@node3 ~]# tar zxvf mongodb-linux-x86_64-2.0.4.tgz -C /usr/local/[root@node3 ~]# mv /usr/local/mongodb-linux-x86_64-2.0.4 /usr/local/mongodb[root@node3 ~]# groupadd -r mongod[root@node3 ~]# useradd -M -r -g mongod -d /mongodb/data -s /bin/false -c mongod mongod[root@node3 ~]# mkdir /mongodb/data/shard13 -p[root@node3 ~]# mkdir /mongodb/data/shard23 -p[root@node3 ~]# mkdir /mongodb/data/config -p[root@node3 ~]# mkdir -p /var/log/mongo[root@node3 ~]# chown -R mongod.mongod /mongodb/data/ /var/log/mongo/
配置mongod
mongod在启动时通过命令行选项或配置文件(如/etc/mongod.conf)接口读取其配置属性。这两种配置接口能够提供相同的功能,因此,管理员可以根据偏好进行选择。如果要通过配置文件读取配置选项,则可以在启动mongod时使用--config或-f选项来指定配置文件的位置。
mongod的基本配置参数
配置mongod的基本工作属性可通过一些常用参数进行。
fork={true|false}: 是否以daemon方式启动mongod,为true表示其启动后其自动转入后台工作;
bind_ip=IP:指定mongod监听的IP地址;
port=PORT:指定mongod监听的端口,默认为27017;
quiet={true|false}:是否工作于静默模式,以最小化的记录日志信息;正常条件下,应该设置为true,仅在调试时才将其设置为false;
dbpath=/PATH/TO/SOMEWHERE:指定mongod存储数据的位置,通常为/data/mongodb、/var/lib/mongodb或/srv/mongodb等;
logpath=/PATH/TO/SOMEFILE:日志文件路径,例如/var/log/mongodb/mongod.log;如果没有指定文件路径,日志信息则会发往标准输出;
logappend={true|false}:设定mongod在启动时是否不覆盖日志文件中的原有日志信息,true表示以追加方式记录日志,即不覆盖;
journal={true|false}:是否启动日志功能;日志功能是保证工作于单实例模式的mongod写入数据持久性的惟一途径;
安全相关的配置参数
数据库服务器的安全机制可有效防范非授权访问。
bind_ip=IP: 指定mongod监听的IP地址;生产环境中,通常需要将其指定为一个可接受外部客户端请求的IP地址,但应该仅开放需要接受客户端请求的IP地址;如果需要指定多个地址,彼此间使用逗号分隔即可;
nounixsocket={true|false}:是否禁用mongodb的Unix套接字功能,默认为启用;主要用于本地通信;
auth={true|false}:是否启用认证功能;如果要启用,远程客户端需要被授权方能访问mongodb服务器;
配置replica sets
[root@node1 ~]# mv /etc/mongod.conf /etc/mongod.conf.bak[root@node1 ~]# vi /etc/mongod.confshardsvr = truereplSet = shard1port = 27017dbpath=/mongodb/data/shard11oplogSize = 100logpath=/var/log/mongo/shard11.loglogappend = truefork = true[root@node2 ~]# mv /etc/mongod.conf /etc/mongod.conf.bak[root@node2 ~]# vi /etc/mongod.confshardsvr = truereplSet = shard1port = 27017dbpath=/mongodb/data/shard12oplogSize = 100logpath=/var/log/mongo/shard12.loglogappend = truefork = true[root@node3 ~]# mv /etc/mongod.conf /etc/mongod.conf.bak[root@node3 ~]# vi /etc/mongod.confshardsvr = truereplSet = shard1port = 27017dbpath=/mongodb/data/shard13oplogSize = 100logpath=/var/log/mongo/shard13.loglogappend = truefork = true
启动replica sets
[root@node1 ~]# /usr/local/mongodb/bin/mongod -f /etc/mongod.conf[root@node1 ~]# netstat -anptl | grep mongtcp 0 0 0.0.0.0:28017 0.0.0.0:* LISTEN 5888/mongodtcp 0 0 0.0.0.0:27017 0.0.0.0:*
同样启动node2,node3的shard1
配置shard2用到的replica sets:
[root@node1 ~]# vi /etc/mongod2.confshardsvr = truereplSet = shard2port = 27018dbpath=/mongodb/data/shard21oplogSize = 100logpath=/var/log/mongo/shard21.loglogappend = truefork = true[root@node2 ~]# vi /etc/mongod2.confshardsvr = truereplSet = shard2port = 27018dbpath=/mongodb/data/shard22oplogSize = 100logpath=/var/log/mongo/shard22.loglogappend = truefork = true[root@node3 ~]# vi /etc/mongod.confshardsvr = truereplSet = shard2port = 27018dbpath=/mongodb/data/shard23oplogSize = 100logpath=/var/log/mongo/shard23.loglogappend = truefork = true
分别启动node1,node2,node3的shard2
[root@node1 ~]# /usr/local/mongodb/bin/mongod -f /etc/mongod2.conf[root@node2 ~]# /usr/local/mongodb/bin/mongod -f /etc/mongod2.conf[root@node3 ~]# /usr/local/mongodb/bin/mongod -f /etc/mongod2.conf
登录其中一个shard1的mongod,初始化复制集
[root@node1 ~]# mongo 192.168.30.116:27017> config = {_id: 'shard1',members: [ {_id: 0, host: '192.168.30.116:27017'}, {_id: 1, host: '192.168.30.117:27017'}, {_id: 2, host: '192.168.30.119:27017'}]}> rs.initiate(config);[root@node1 ~]# mongo 192.168.30.116:27018> config = {_id: 'shard2',members: [ {_id: 0, host: '192.168.30.116:27018'}, {_id: 1, host: '192.168.30.117:27018'}, {_id: 2, host: '192.168.30.119:27018'}]}> rs.initiate(config);
到此就配置好了二个replica sets,也就是准备好了二个shards
配置三台config server
[root@node1 ~]# vi /etc/config.confconfigsvr = trueport = 20000dbpath=/mongodb/data/configlogpath=/var/log/mongo/config.loglogappend = truefork = true[root@node2 ~]# vi /etc/config.confconfigsvr = trueport = 20000dbpath=/mongodb/data/configlogpath=/var/log/mongo/config.loglogappend = truefork = true[root@node3 ~]# vi /etc/config.confconfigsvr = trueport = 20000dbpath=/mongodb/data/configlogpath=/var/log/mongo/config.loglogappend = truefork = true[root@node1 ~]# /usr/local/mongodb/bin/mongod -f /etc/config.conf[root@node2 ~]# /usr/local/mongodb/bin/mongod -f /etc/config.conf[root@node3 ~]# /usr/local/mongodb/bin/mongod -f /etc/config.conf
配置启动mongs
[root@node1 ~]# /usr/local/mongodb/bin/mongos --configdb 192.168.30.116:20000,192.168.30.117:20000,192.168.30.119:20000 --port 30000 --chunkSize 5 --logpath /mongodb/data/mongos.log --logappend --fork[root@node2 ~]# /usr/local/mongodb/bin/mongos --configdb 192.168.30.116:20000,192.168.30.117:20000,192.168.30.119:20000 --port 30000 --chunkSize 5 --logpath /mongodb/data/mongos.log --logappend --fork[root@node3 ~]# /usr/local/mongodb/bin/mongos --configdb 192.168.30.116:20000,192.168.30.117:20000,192.168.30.119:20000 --port 30000 --chunkSize 5 --logpath /mongodb/data/mongos.log --logappend --fork
查看是mongod,config,mongos是否启动
[root@node1 ~]# ss -antpl | grep mongLISTEN 0 128 192.168.30.116:30000 *:* users:(("mongos",3627,14))LISTEN 0 128 192.168.30.116:28017 *:* users:(("mongod",22036,6))LISTEN 0 128 192.168.30.116:28018 *:* users:(("mongod",3287,8))LISTEN 0 128 192.168.30.116:31000 *:* users:(("mongos",3627,13))LISTEN 0 128 192.168.30.116:20000 *:* users:(("mongod",3487,6))LISTEN 0 128 192.168.30.116:21000 *:* users:(("mongod",3487,7))LISTEN 0 128 192.168.30.116:27017 *:* users:(("mongod",22036,7))LISTEN 0 128 192.168.30.116:27018 *:* users:(("mongod",3287,7))
都已启动成功
连接到其中一个mongos进程,并切换到admin数据库做以下配置:
[root@node1 ~]# /usr/local/mongodb/bin/mongo 192.168.30.116:30000/adminMongoDB shell version: 2.0.4connecting to: 192.168.30.116:30000/admin>db.runCommand( { addshard: "shard1/192.168.30.116:27017,192.168.30.117:27017,192.168.30.119:27017″,name:"s1",maxsize:20480});>db.runCommand( { addshard: "shard2/192.168.30.116:27018,192.168.30.117:27018,192.168.30.119:27018″,name:"s2",maxsize:20480});
查看加入的2个shard
mongos> db.runCommand( { listshards : 1 } );{ "shards" : [ { "_id" : "s1", "host" : "shard1/192.168.30.116:27017,192.168.30.117:27017,192.168.30.119:27017" }, { "_id" : "s2", "host" : "shard2/192.168.30.116:27018,192.168.30.117:27018,192.168.30.119:27018" } ], "ok" : 1}
启动用数据库分片
mongos> db.runCommand({enablesharding : "testdb"});{ "ok" : 1 }
指定分片的key
mongos> db.runCommand({shardcollection :"testdb.user",key :{_id:1}});{ "collectionsharded" : "test.user", "ok" : 1 }
插入20W行数据测试是否分片
mongos> use testdbswitched to db testdbmongos> for(var i=1;i<=200000;i++) db.user.insert({age:i,name:"ljl",addr:"beijing",country:"china"})mongos> db.user.stats(){ "sharded" : true, "flags" : 1, "ns" : "testdb.user", "count" : 70001, "numExtents" : 7, "size" : 6160096, "storageSize" : 11190272, "totalIndexSize" : 2289280, "indexSizes" : { "_id_" : 2289280 }, "avgObjSize" : 88.00011428408166, "nindexes" : 1, "nchunks" : 10, "shards" : { "s1" : { "ns" : "testdb.user", "count" : 0, "size" : 0, "storageSize" : 8192, "numExtents" : 1, "nindexes" : 1, "lastExtentSize" : 8192, "paddingFactor" : 1, "flags" : 1, "totalIndexSize" : 8176, "indexSizes" : { "_id_" : 8176 }, "ok" : 1 }, "s2" : { "ns" : "testdb.user", "count" : 70001, "size" : 6160096, "avgObjSize" : 88.00011428408166, "storageSize" : 11182080, "numExtents" : 6, "nindexes" : 1, "lastExtentSize" : 8388608, "paddingFactor" : 1, "flags" : 1, "totalIndexSize" : 2281104, "indexSizes" : { "_id_" : 2281104 }, "ok" : 1 } }, "ok" : 1}
以上信息发现数据已经分布在不同分片上
下面模拟自动故障转移功能,在node1关闭shard1
[root@node1 ~]# /usr/local/mongodb/bin/mongod --shutdown --shardsvr --replSet shard1 --port 27017 --dbpath /mongodb/data/shard11killing process with pid: 22036[root@node1 ~]# /usr/local/mongodb/bin/mongo 192.168.30.116:30000/testdbMongoDB shell version: 2.0.4connecting to: 192.168.30.116:30000/testdbmongos> db.user.find(){ "_id" : ObjectId("5338092034594e87a8cb007b"), "age" : 1, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb007c"), "age" : 2, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb007d"), "age" : 3, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb007e"), "age" : 4, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb007f"), "age" : 5, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0080"), "age" : 6, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0081"), "age" : 7, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0082"), "age" : 8, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0083"), "age" : 9, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0084"), "age" : 10, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0085"), "age" : 11, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0086"), "age" : 12, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0087"), "age" : 13, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0088"), "age" : 14, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb0089"), "age" : 15, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb008a"), "age" : 16, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb008b"), "age" : 17, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb008c"), "age" : 18, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb008d"), "age" : 19, "name" : "ljl", "addr" : "beijing", "country" : "china" }{ "_id" : ObjectId("5338092034594e87a8cb008e"), "age" : 20, "name" : "ljl", "addr" : "beijing", "country" : "china" }has moremongos>
以上信息发现数据一切正常
分片注意的几个事项:
选择sharding key的标准:
应该在哪儿存储数据?
应该从哪儿得到希望的数据?
基本法则:
sharding key应该是主键;
sharding key应该能尽量保证避免跨分片查询。