Mongo官网——Sharding

官方介绍

Sharding（分片） is a method for distributing data across multiple machines. MongoDB uses sharding to support deployments with very large data sets and high throughput operations.

个人理解：
  • 分片是一种支持海量数据存储并进行高吞吐量操作的方式
  • 在大数据集和高吞吐量操作的情况下，对单一的服务器硬件要求较高（一般要求算力优异的CPU提供运算能力，RAM或者DISK也要足够大）。传统的方式就是对服务器硬件进行升级，而这样的做的成本往往很高（垂直扩展）。而MongoDB提供的分片，其实上就是使用多数的廉价服务器构建成集群，提供海量的数据存储以及并行计算的能力（水平扩展）

一、 原理图

mongodb分片集群状态 mongodb分片集群原理_MongoDB

二、 分片集群的组件

  • shard server：用于存储实际的数据块，实际生产环境中一个shard server角色可由几台机器组个一个replica set承担，防止主机单点故障
  • config server：顾名思义为配置服务器，存储所有数据库元信息（路由、分片）的配置。
  • mongos server：数据库集群请求的入口，所有的请求都通过mongos进行协调，不需要在应用程序添加一个路由选择器，mongos自己就是一个请求分发中心，它负责把对应的数据请求请求转发到对应的shard服务器上。在生产环境通常有多mongos作为请求的入口，防止其中一个挂掉所有的mongodb请求都没有办法操作。

三、 搭建步骤

注意：3.4版本后，config server需要搭建集群

1. 准备5台服务器
shard1 192.168.128.156:28001
shard2 192.168.128.156:28002
shard3 192.168.128.156:28003
config1 192.168.128.156:28004
config2 192.168.128.156:28005
config3 192.168.128.156:28006
mongos 192.168.128.156:28007
2. 启动shard服务器
mkdir -p /data/{shard1,shard2,shard3}
mkdir -p /data/{config1,config2,config3}
mkdir -p /data/mongos
mongodb/bin/mongod --port 28001 --dbpath=/data/shard1/ --bind_ip 192.168.128.156 --shardsvr
mongodb/bin/mongod --port 28002 --dbpath=/data/shard2/ --bind_ip 192.168.128.156 --shardsvr
mongodb/bin/mongod --port 28003 --dbpath=/data/shard3/ --bind_ip 192.168.128.156 --shardsvr
3. 配置启动config服务器
# 启动
mongodb/bin/mongod --port 28004 --dbpath=/data/config1/ --bind_ip 192.168.128.156  --configsvr --replSet rs
mongodb/bin/mongod --port 28005 --dbpath=/data/config2/ --bind_ip 192.168.128.156  --configsvr --replSet rs
mongodb/bin/mongod --port 28006 --dbpath=/data/config3/ --bind_ip 192.168.128.156  --configsvr --replSet rs

# 配置副本集
rs.initiate( {
 _id : "rs",
 members: [
    { _id: 0, host: "192.168.128.156:28004" },
    { _id: 1, host: "192.168.128.156:28005" },
    { _id: 2, host: "192.168.128.156:28006" }
 ]
})

# 查看配置服务器副本集状态
rs:SECONDARY> rs.status();
{
	"set" : "rs",
	"date" : ISODate("2018-01-19T01:42:20.535Z"),
	"myState" : 1,
	"term" : NumberLong(1),
	"configsvr" : true,
	"heartbeatIntervalMillis" : NumberLong(2000),
	"optimes" : {
		"lastCommittedOpTime" : {
			"ts" : Timestamp(1516326131, 1),
			"t" : NumberLong(1)
		},
		"readConcernMajorityOpTime" : {
			"ts" : Timestamp(1516326131, 1),
			"t" : NumberLong(1)
		},
		"appliedOpTime" : {
			"ts" : Timestamp(1516326131, 1),
			"t" : NumberLong(1)
		},
		"durableOpTime" : {
			"ts" : Timestamp(1516326131, 1),
			"t" : NumberLong(1)
		}
	},
	"members" : [
		{
			"_id" : 0,
			"name" : "192.168.128.156:28004",
			"health" : 1,
			"state" : 1,
			"stateStr" : "PRIMARY",
			"uptime" : 175,
			"optime" : {
				"ts" : Timestamp(1516326131, 1),
				"t" : NumberLong(1)
			},
			"optimeDate" : ISODate("2018-01-19T01:42:11Z"),
			"electionTime" : Timestamp(1516326018, 1),
			"electionDate" : ISODate("2018-01-19T01:40:18Z"),
			"configVersion" : 1,
			"self" : true
		},
		{
			"_id" : 1,
			"name" : "192.168.128.156:28005",
			"health" : 1,
			"state" : 2,
			"stateStr" : "SECONDARY",
			"uptime" : 133,
			"optime" : {
				"ts" : Timestamp(1516326131, 1),
				"t" : NumberLong(1)
			},
			"optimeDurable" : {
				"ts" : Timestamp(1516326131, 1),
				"t" : NumberLong(1)
			},
			"optimeDate" : ISODate("2018-01-19T01:42:11Z"),
			"optimeDurableDate" : ISODate("2018-01-19T01:42:11Z"),
			"lastHeartbeat" : ISODate("2018-01-19T01:42:18.777Z"),
			"lastHeartbeatRecv" : ISODate("2018-01-19T01:42:19.933Z"),
			"pingMs" : NumberLong(0),
			"syncingTo" : "192.168.128.156:28004",
			"configVersion" : 1
		},
		{
			"_id" : 2,
			"name" : "192.168.128.156:28006",
			"health" : 1,
			"state" : 2,
			"stateStr" : "SECONDARY",
			"uptime" : 133,
			"optime" : {
				"ts" : Timestamp(1516326131, 1),
				"t" : NumberLong(1)
			},
			"optimeDurable" : {
				"ts" : Timestamp(1516326131, 1),
				"t" : NumberLong(1)
			},
			"optimeDate" : ISODate("2018-01-19T01:42:11Z"),
			"optimeDurableDate" : ISODate("2018-01-19T01:42:11Z"),
			"lastHeartbeat" : ISODate("2018-01-19T01:42:18.778Z"),
			"lastHeartbeatRecv" : ISODate("2018-01-19T01:42:20.030Z"),
			"pingMs" : NumberLong(0),
			"syncingTo" : "192.168.128.156:28004",
			"configVersion" : 1
		}
	],
	"ok" : 1,
	"operationTime" : Timestamp(1516326131, 1),
	"$gleStats" : {
		"lastOpTime" : Timestamp(1516326007, 1),
		"electionId" : ObjectId("7fffffff0000000000000001")
	},
	"$clusterTime" : {
		"clusterTime" : Timestamp(1516326131, 1),
		"signature" : {
			"hash" : BinData(0,"AAAAAAAAAAAAAAAAAAAAAAAAAAA="),
			"keyId" : NumberLong(0)
		}
	}
}

mongodb分片集群状态 mongodb分片集群原理_MongoDB_02

4. 启动mongos服务器
mongodb/bin/mongos --port 28007 --bind_ip 192.168.128.156 --configdb rs/192.168.128.156:28004,192.168.128.156:28005,192.168.128.156:28006
5. 初始化分片集群
# 使用客户端命令连接分片集群
mongodb/bin/mongo 192.168.128.156:28007

# 设置chunk大小
use config
db.settings.save( { _id:"chunksize", value: 1 } )

# 添加分片节点
db.runCommand({addShard:"192.168.128.156:28001"});
db.runCommand({addShard:"192.168.128.156:28002"});
db.runCommand({addShard:"192.168.128.156:28003"});
mongodb/bin/mongo 192.168.128.156:28007

# MongoDB分片是针对集合的，要想使集合支持分片，首先需要使其数据库支持分片，为数据库testdb启动分片
sh.enableSharding("testdb");

# 为分片字段建立索引，同时为集合指定片键
use testdb
db.users.ensureIndex({name:1});

# 启用集合分片，为其指定片键
sh.shardCollection("testdb.users",{name:1});

mongodb分片集群状态 mongodb分片集群原理_NOSQL_03


mongodb分片集群状态 mongodb分片集群原理_mongodb分片集群状态_04

6. 测试
// 连接mongos，插入50W数据测试下分片
for(var i = 0;i<500000;i++){
	db.users.insert({"name":"zs"+i,"age":i});
}
mongos> sh.status();
--- Sharding Status --- 
  sharding version: {
  	"_id" : 1,
  	"minCompatibleVersion" : 5,
  	"currentVersion" : 6,
  	"clusterId" : ObjectId("5a6163a5130c3601a3a20db4")
  }
  shards:
        {  "_id" : "shard0000",  "host" : "192.168.128.156:28001",  "state" : 1 }
        {  "_id" : "shard0001",  "host" : "192.168.128.156:28002",  "state" : 1 }
        {  "_id" : "shard0002",  "host" : "192.168.128.156:28003",  "state" : 1 }
  active mongoses:
        "3.6.2" : 1
  autosplit:
        Currently enabled: yes
  balancer:
        Currently enabled:  yes
        Currently running:  yes
        Collections with active migrations: 
                testdb.users started at Fri Jan 19 2018 11:25:25 GMT+0800 (CST)
        Failed balancer rounds in last 5 attempts:  0
        Migration Results for the last 24 hours: 
                5 : Success
  databases:
        {  "_id" : "config",  "primary" : "config",  "partitioned" : true }
                config.system.sessions
                        shard key: { "_id" : 1 }
                        unique: false
                        balancing: true
                        chunks:
                                shard0000	1
                        { "_id" : { "$minKey" : 1 } } -->> { "_id" : { "$maxKey" : 1 } } on : shard0000 Timestamp(1, 0) 
        {  "_id" : "testdb",  "primary" : "shard0000",  "partitioned" : true }
                testdb.users
                        shard key: { "name" : 1 }
                        unique: false
                        balancing: true
                        chunks:
                                shard0000	3
                                shard0001	3
                                shard0002	5
                        { "name" : { "$minKey" : 1 } } -->> { "name" : "zs1" } on : shard0002 Timestamp(5, 0) 
                        { "name" : "zs1" } -->> { "name" : "zs108900" } on : shard0002 Timestamp(6, 2) 
                        { "name" : "zs108900" } -->> { "name" : "zs17318" } on : shard0002 Timestamp(6, 3) 
                        { "name" : "zs17318" } -->> { "name" : "zs19072" } on : shard0002 Timestamp(6, 4) 
                        { "name" : "zs19072" } -->> { "name" : "zs28146" } on : shard0001 Timestamp(5, 1) 
                        { "name" : "zs28146" } -->> { "name" : "zs42" } on : shard0001 Timestamp(4, 3) 
                        { "name" : "zs42" } -->> { "name" : "zs5163" } on : shard0001 Timestamp(4, 4) 
                        { "name" : "zs5163" } -->> { "name" : "zs60703" } on : shard0002 Timestamp(6, 0) 
                        { "name" : "zs60703" } -->> { "name" : "zs6978" } on : shard0000 Timestamp(6, 1) 
                        { "name" : "zs6978" } -->> { "name" : "zs8724" } on : shard0000 Timestamp(5, 4) 
                        { "name" : "zs8724" } -->> { "name" : { "$maxKey" : 1 } } on : shard0000 Timestamp(1, 3)

分片结果

mongodb分片集群状态 mongodb分片集群原理_MongoDB_05


mongodb分片集群状态 mongodb分片集群原理_Sharding_06

四、常见问题

MongoDB在分片后的集合上进行db.collection.count()操作时，出现结果不准确的现象，需要采用聚合的方法获取集合的count结果。

参考资料： https://docs.mongodb.com/manual/reference/method/db.collection.count/

db.collection.aggregate( [
   { $count: "myCount" }
])

或

db.collection.aggregate( [
   { $group: { _id: null, count: { $sum: 1 } } }
   { $project: { _id: 0 } }
] )