MongoDB集群部署：Replic Set + Sharding
一、方案：Replic Set + Sharding
1)虚拟机环境：
三台centos6.6虚拟机（参考http://blog.****.net/chxr1620/article/details/76695682）
2)mongodb的版本：mongodb-linux-x86_64-2.6.12.tar
二、安装部署
1. tar zxvf mongodb-linux-x86_64-2.x.x.tgz
2. rm -rf /opt/mongodbmv mongodb-linux-x86_64-2.x.x/ /opt/mongodb
3. 自动部署脚本 ,分别在centos1,centos2,centos3进行执行
该脚本（mongodb_deploy.sh）内容为如下：
#!/bin/sh
echo"Plear input the Centos's index(1 or 2 or 3)?"
read cindex
rm-rf /work/mongodb
mkdir -p /work/mongodb/shard1$cindex
cat> /work/mongodb/shard1$cindex.conf << EOF
shardsvr=true
replSet=shard1
port=28017
dbpath=/work/mongodb/shard1$cindex
oplogSize=500
logpath=/work/mongodb/shard1$cindex.log
logappend=true
fork=true
rest=true
nojournal=true
EOF
mkdir -p /work/mongodb/shard2$cindex
cat> /work/mongodb/shard2$cindex.conf << EOF
shardsvr=true
replSet=shard2
port=28018
dbpath=/work/mongodb/shard2$cindex
oplogSize=500
logpath=/work/mongodb/shard2$cindex.log
logappend=true
fork=true
rest=true
nojournal=true
EOF
mkdir-p /work/mongodb/config
cat > /work/mongodb/config$cindex.conf << EOF
configsvr=true
dbpath=/work/mongodb/config/
port=20000
logpath=/work/mongodb/config$cindex.log
logappend=true
fork=true
nojournal=true
EOF
mkdir-p /work/mongodb/arbiter1
cat> /work/mongodb/arbiter1.conf << EOF
shardsvr=true
replSet=shard1
port=28031
dbpath=/work/mongodb/arbiter1
oplogSize=100
logpath=/work/mongodb/arbiter1.log
logappend=true
fork=true
rest=true
nojournal=true
EOF
echo"finished arbiter1...."
mkdir-p /work/mongodb/arbiter2
cat> /work/mongodb/arbiter2.conf << EOF
shardsvr=true
replSet=shard2
port=28032
dbpath=/work/mongodb/arbiter2
oplogSize=100
logpath=/work/mongodb/arbiter2.log
logappend=true
fork=true
rest=true
nojournal=true
EOF
echo"finished arbiter2....."
mkdir-p /work/mongodb/mongos$cindex
cat> /work/mongodb/mongos$cindex.conf << EOF
configdb=centos1:20000,centos2:20000,centos3:20000
port=28885
chunkSize=100
logpath=/work/mongodb/mongos$cindex.log
logappend=true
fork=true
EOF
echo"finished mongos........"
echo"begin shard1"
/opt/mongodb/bin/mongod--config /work/mongodb/shard1$cindex.conf
echo"begin shard2"
/opt/mongodb/bin/mongod--config /work/mongodb/shard2$cindex.conf
echo"begin arbiter1"
/opt/mongodb/bin/mongod--config /work/mongodb/arbiter1.conf
echo"begin arbiter2"
/opt/mongodb/bin/mongod--config /work/mongodb/arbiter2.conf
echo"begin config"
/opt/mongodb/bin/mongod--config /work/mongodb/config$cindex.conf
4. 各个虚拟机上分别执行,开启mongos服务，脚本(mongos_deploy.sh)内容如下：
#!/bin/bash
echo"the current Centos' index (1 or 2 or 3)?"
read cindex
/opt/mongodb/bin/mongos --config/work/mongodb/mongos$cindex.conf
5. 配置分片副本集
任意登陆一个机器，如centos1,设置第一个分片副本集
/opt/mongodb/bin/mongo centos1:28017/admin
定义副本集配置：
config = { _id:"shard1", members:[ { _id:0,host:"centos1:28017"}, { _id:1,host:"centos2:28017"}, {_id:2,host:"centos3:28017",slaveDelay:7200,priority:0}, {_id:3,host:"centos1:28031",arbiterOnly:true}, {_id:4,host:"centos2:28031",arbiterOnly:true}, {_id:5,host:"centos3:28031",arbiterOnly:true} ] };
初始化副本集配置。
rs.initiate(config):
注：配置分片信息时遇到的报错（有如下几种）
{
"errmsg" : "exception:Can't take a write lock while out of disk space",
"code" : 14031,
"ok" : 0
}
{
"ok" : 0,
"errmsg" : "couldn'tinitiate : cmdline oplogsize (2048) different than existing (0) see:http://dochub.mongodb.org/core/increase-oplog"
}
以上两种情况，重新从上述第3步开始执行。
{"ok" : 0, "errmsg" : "couldn't initiate : new fileallocation failure" }
对于这个错误信息，主要是系统的存储空间过少问题造成的，可以修改第3步中oplogSize的大小，然后重新执行从第3步开始执行。
同样的设置第二个分片集
/opt/mongodb/bin/mongo centos1:28018/admin
config = { _id:"shard2", members:[ { _id:0,host:"centos1:28018"}, { _id:1,host:"centos2:28018"}, { _id:2,host:"centos3:28018",slaveDelay:7200,priority:0}, {_id:3,host:"centos1:28032",arbiterOnly:true}, { _id:4,host:"centos2:28032",arbiterOnly:true}, {_id:5,host:"centos3:28032",arbiterOnly:true} ] };
rs.initiate(config)
6. 添加分片
进入任意一台虚拟机
/opt/mongodb/bin/mongo centos1:2885/admin
串联路由服务器与分配副本集1,2
db.runCommand({“addshard”:”shard1/centos:28017,centos2:28017”})
db.runCommand({“addshard”:”shard2/centos:28018,centos2:28018”})
因为centos3:28017,centos3:28018是延迟节点，所以不在副本集当中。
查看分片服务器的配置：
db.runCommand({listsshards:1});
指定NCDB库分片生效
db.runCommand({“enablesharding”:”NCDB”})
指定分片集合与片键
db.runCommand({shardcollection:”NCDB.test1”,key:{id:1}})
7. 测试分片结果
下面我们进一步对分片结果进行测试
#连接mongos服务器并对库表分片
[[email protected] ~]$ mongoracdb:28885/admin
MongoDB shell version: 3.2.3
connecting to:racdb:28885/admin
mongos> db.runCommand( {enablesharding:"testdb"});
{ "ok" : 1 }
mongos> db.runCommand( {shardcollection: "testdb.table1",key : {id: 1} } )
{ "collectionsharded":"testdb.table1", "ok" : 1 }
#插入测试数据
for (var i = 1; i <= 10000;i++)
db.table1.save({id:i,"test1":"licz"});
进入任意一台虚拟机
/opt/mongodb/bin/mongo centos1:28885/admin
db.runCommand({enablesharding:”testdb”})
指定分片库时的错误参考：
https://stackoverflow.com/questions/19918956/can-not-create-shards-sharding-not-enabled-for-db
参考：http://blog.****.net/lichangzai/article/details/50927588