随着大数据海量数据的不断涌现，分布式，横向扩展是系统扩展的重要方式之一。基于文档的NoSQL领头羊mongoDB正是这样一个分布式系统，通过分片集群将所有数据分成数据段，并将每个数据段写入不同的节点。本文简要描述mongoDB分片特性，以及给出演示，快速体验mongoDB分片方式。
一、为什么需要shard
存储容量需求超出单机磁盘容量
活跃的数据集超出单机内存容量，导致很多请求都要从磁盘读取数据，影响性能
写IOPS超出单个mongoDB节点的写服务能力
mongoDB支持自动分片以及手动分片，分片的基本单位是集合
二、Shared cluster架构
Mongos
客户端访问路由节点，mongos进行数据读写
Config Server
保存元数据以及集群配置信息
Shard Server
每一个shard包含特定集合数据的一部分，且shard可以配置为复制集
三、什么是主分片
主分片用于存储所有未开启分片集合的数据
每一个数据库都有一个主分片
通过movePrimary命令改变主分片
基于已经使用了复制集的环境，在开启一个分片集群的情形下，已经存在的数据库依旧位于原有的分片
可以创建指向单个分片的片键
四、演示mongodb分片
//演示环境，在同一台机器上使用不同的端口来实现
//包含2个分片节点，端口号(28000,28001)，一个config server，端口号(29000)
//本演示未涉及到副本集
# more /etc/redhat-release
CentOS release 6.7 (Final)
# mongod --version
db version v3.0.12
git version: 33934938e0e95d534cebbaff656cde916b9c3573
//创建相应目录
# mkdir -pv /data/{s1,s2,con1}
//以下分别启动2个mongod分片实例
# mongod --shardsvr --dbpath /data/s1 --port 28000 --logpath /data/s1/s1.log --smallfiles --oplogSize 128 --fork
# mongod --shardsvr --dbpath /data/s2 --port 28001 --logpath /data/s2/s2.log --smallfiles --oplogSize 128 --fork
//以下分别启动config server以及mongos
# mongod --configsvr --dbpath /data/con1 --port 29000 --logpath /data/con1/config.log --fork
# mongos --configdb localhost:29000 --logpath /var/log/mongos.log --fork
//登陆到mongo客户端
# mongo
MongoDB shell version: 3.0.12
connecting to: test
mongos> show dbs;
admin (empty)
config 0.016GB
//添加分片
mongos> sh.addShard("localhost:28000")
{ "shardAdded" : "shard0000", "ok" : 1 }
mongos> sh.addShard("localhost:28001")
{ "shardAdded" : "shard0001", "ok" : 1 }
//插入文档
mongos> db
test
mongos> for (i=0;i<1000000;i++){
db.users.insert(
{
i:i,
username:"user"+i,
age:Math.floor(Math.random()*120),
created:new Date()
}
);
}
WriteResult({ "nInserted" : 1 })
//查看分片状态
mongos> sh.status()
--- Sharding Status ---
sharding version: {
"_id" : 1,
"minCompatibleVersion" : 5,
"currentVersion" : 6,
"clusterId" : ObjectId("57c64b0e0ea3f71b79b3fb8e")
}
shards:
{ "_id" : "shard0000", "host" : "localhost:28000" }
{ "_id" : "shard0001", "host" : "localhost:28001" }
balancer:
Currently enabled: yes
Currently running: no
Failed balancer rounds in last 5 attempts: 0
Migration Results for the last 24 hours:
No recent migrations
databases:
{ "_id" : "admin", "partitioned" : false, "primary" : "config" }
{ "_id" : "test", "partitioned" : false, "primary" : "shard0000" }
//通过sh.status()可知，当前的集群环境有2个分片为shard0000，shard0001，在当前主机的28000以及28001端口
//当前有2个数据库，一个是admin，一个test数据库
//可以看到当前我们刚刚创建的集合users位于test库,shard0000，即当前shard0000为主分片
//要对一个集合分片，需要先开启库级分片，如下
mongos> sh.enableSharding("test")
{ "ok" : 1 }
mongos> sh.status()
--- Sharding Status ---
sharding version: {
.........
databases:
{ "_id" : "admin", "partitioned" : false, "primary" : "config" }
{ "_id" : "test", "partitioned" : true, "primary" : "shard0000" }
//在上面的查询中可以看出test数据库partitioned的value已经变成了true
//为集合users添加索引，sharding会根据索引将其数据分布到不同的片上，所以索引是必须的
//这个sharding与RDBMS的分区表的实质一样
//RDBMS的分区表是将数据分布在单台机器的多个磁盘上
//MongoDB则是将数据分布在不同机器的不同磁盘上
mongos> db.users.ensureIndex({username:1})
{
"raw" : {
"localhost:28000" : {
"createdCollectionAutomatically" : false,
"numIndexesBefore" : 1,
"numIndexesAfter" : 2,
"ok" : 1
}
},
"ok" : 1
}
//查看已经创建的索引
mongos> db.users.getIndexes()
[
{
"v" : 1,
"key" : {
"_id" : 1
},
"name" : "_id_",
"ns" : "test.users"
},
{
"v" : 1,
"key" : {
"username" : 1
},
"name" : "username_1",
"ns" : "test.users"
}
]
//开启集合分片
mongos> sh.shardCollection("test.users",{"username":1})
{ "collectionsharded" : "test.users", "ok" : 1 }
//查看集合的状态
mongos> sh.status()
--- Sharding Status ---
sharding version: {
"_id" : 1,
"minCompatibleVersion" : 5,
"currentVersion" : 6,
"clusterId" : ObjectId("57c689d2425d38d84702dbf4")
}
shards:
{ "_id" : "shard0000", "host" : "localhost:28000" }
{ "_id" : "shard0001", "host" : "localhost:28001" }
balancer:
Currently enabled: yes
Currently running: no
Failed balancer rounds in last 5 attempts: 0
Migration Results for the last 24 hours:
10 : Success
1 : Failed with error 'chunk too big to move', from shard0000 to shard0001
databases:
{ "_id" : "admin", "partitioned" : false, "primary" : "config" }
{ "_id" : "test", "partitioned" : true, "primary" : "shard0000" }
test.users
shard key: { "username" : 1 }
chunks:
shard0000 4
shard0001 3
{ "username" : { "$minKey" : 1 } } -->> { "username" : "user167405" } on : shard0001 Timestamp(2, 0)
{ "username" : "user167405" } -->> { "username" : "user234814" } on : shard0001 Timestamp(3, 0)
{ "username" : "user234814" } -->> { "username" : "user302222" } on : shard0001 Timestamp(4, 0)
{ "username" : "user302222" } -->> { "username" : "user369631" } on : shard0000 Timestamp(4, 1)
{ "username" : "user369631" } -->> { "username" : "user639266" } on : shard0000 Timestamp(1, 1)
{ "username" : "user639266" } -->> { "username" : "user908900" } on : shard0000 Timestamp(1, 2)
{ "username" : "user908900" } -->> { "username" : { "$maxKey" : 1 } } on : shard0000 Timestamp(1, 3)
//从上面的结果可知，基于username分片，分成了7个片，或者说7个chunk,其中shard0000 4个，shard0001 3个
//$minKey，$maxKey表示最小值或最大值，也可以理解为负无穷和正无穷。
//$minKey到user167405位于shard0001，user167405到user234814位于shard0001，其余依此类推
//查看集合的状态
mongos> db.users.stats()
{
"sharded" : true,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0.
It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false, //Author : Leshami
"ns" : "test.users", //Blog : http://blog.csdn.net/leshami
"count" : 1000000,
"numExtents" : 20,
"size" : 112000000,
"storageSize" : 212533248,
"totalIndexSize" : 66863328,
"indexSizes" : {
"_id_" : 32916576,
"username_1" : 33946752
},
"avgObjSize" : 112,
"nindexes" : 2,
"nchunks" : 7,
"shards" : {
"shard0000" : {
"ns" : "test.users",
"count" : 775304,
"size" : 86834048,
"avgObjSize" : 112,
"numExtents" : 12,
"storageSize" : 174735360,
"lastExtentSize" : 50798592,
"paddingFactor" : 1,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0.
It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"nindexes" : 2,
"totalIndexSize" : 47020176,
"indexSizes" : {
"_id_" : 25214784,
"username_1" : 21805392
},
"ok" : 1
},
"shard0001" : {
"ns" : "test.users",
"count" : 224696,
"size" : 25165952,
"avgObjSize" : 112,
"numExtents" : 8,
"storageSize" : 37797888,
"lastExtentSize" : 15290368,
"paddingFactor" : 1,
"paddingFactorNote" : "paddingFactor is unused and unmaintained in 3.0.
It remains hard coded to 1.0 for compatibility only.",
"userFlags" : 1,
"capped" : false,
"nindexes" : 2,
"totalIndexSize" : 19843152,
"indexSizes" : {
"_id_" : 7701792,
"username_1" : 12141360
},
"ok" : 1
}
},
"ok" : 1
}
//上面的结果中可以看到不同shard上对应的记录数，如shard0000为775304，shard0001为224696
//也可以看到很对与当前集合相关的信息，如大小，平均队列长度，索引等等
//下面测试文档查找
mongos> db.users.find({username:"user1000"},{_id:0})
{ "i" : 1000, "username" : "user1000", "age" : 87, "created" : ISODate("2016-08-31T08:31:28.233Z") }
mongos> db.users.find({username:"user639266"},{_id:0})
{ "i" : 639266, "username" : "user639266", "age" : 6, "created" : ISODate("2016-08-31T08:38:30.122Z") }
mongos> db.users.count()
1000000
//以下查看执行计划
//从执行计划可知，当前查询到了分片shard0001，即实现了基于分片的隔离
mongos> db.users.find({username:"user1000"}).explain()
{
"queryPlanner" : {
"mongosPlannerVersion" : 1,
"winningPlan" : {
"stage" : "SINGLE_SHARD",
"shards" : [
{
"shardName" : "shard0001",
"connectionString" : "localhost:28001",
"serverInfo" : {
"host" : "node1.edq.com",
"port" : 28001,
"version" : "3.0.12",
"gitVersion" : "33934938e0e95d534cebbaff656cde916b9c3573"
},
"plannerVersion" : 1,
"namespace" : "test.users",
"indexFilterSet" : false,
"parsedQuery" : {
"username" : {
"$eq" : "user1000"
}
},
"winningPlan" : {
"stage" : "FETCH",
"inputStage" : {
"stage" : "SHARDING_FILTER",
"inputStage" : {
"stage" : "IXSCAN",
"keyPattern" : {
"username" : 1
},
"indexName" : "username_1",
"isMultiKey" : false,
"direction" : "forward",
"indexBounds" : {
"username" : [
"[\"user1000\", \"user1000\"]"
]
}
}
}
},
"rejectedPlans" : [ ]
}
]
}
},
"ok" : 1
} 五、小结
1、mongodb分片的实质是将数据分散到不同的物理机器，以分散IO，提供并发与吞吐量
2、mongodb分片依赖于片键，即任意一个需要开启的集合都需要创建索引
3、开启分片的集合需要首先在DB级别启用库级分片
4、mongodb的分片由分片服务器，配置服务器以及路由服务器组成
5、基于分片可以结合副本集(replicate set)来实现高可用
