mongodb replica set

之前主要介绍了RS(replica set)的安装配置，这里再介绍一下日常的管理，比如添加以及删除节点。

故障转移，这个无疑是RS的重要特性：

<wbr style="line-height:25px"></wbr>

在之前建立好的replica set，直接kill掉进程，剩下的节点就会选举出一个节点作为主节点。

添加删除节点：

删除节点：rs.remove()即可

例如：rs.remove("localhost:20004")

添加节点：mkdir -p /mongo/data4

echo "this is a key for RS">/mongo/key/k4

touch /mongo/mongo_rs_log/rs4.log

建立好相关目录以及文件

mongod --replSet myrs/localhost:20004 --keyFile /mongo/key/k4 --fork --port 20004 --dbpath /mongo/data4/ --logpath=/mongo/mongo_rs_log/rs4.log //启动第四个节点

rs.add({_id:3,host:"localhost:20004"})

通过rs.status()查看状态。

以上是通过oplog来添加节点

其实这里有一个缺点，oplog是copped collections，所以他会循环写，根据lru覆盖数据，所以可能造成和主库数据的不一致，这里mongodb提供了快照加oplog方式，以保持新加节点的一致性，列出具体步骤：

1.一个复制集成员的物理文件来初始化数据，也就是直接copy一个dbpath，建立相应的Key文件，以及日志文件。

[mongo@172_16_3_216 mongo]$ mkdir data5
[mongo@172_16_3_216 mongo]$ cp data1/* data5/

[mongo@172_16_3_216 mongo_rs_log]$ touch rs5.log

[mongo@172_16_3_216key]$cp k1 k5

2.插入验证数据

myrs:PRIMARY> db.bank.insert({computer:"mac"})

3.启动新节点

mongod --replSet myrs/localhost:20005 --keyFile /mongo/key/k5 --fork --port 20005 --dbpath /mongo/data5/ --logpath=/mongo/mongo_rs_log/RS5.log--fastsync

4.添加节点

myrs:PRIMARY> rs.add("localhost:20005")

5.验证数据登录该节点

rs.status()
{
"set" : "myrs",
"date" : ISODate("2011-11-16T09:50:29Z"),
"myState" : 1,
"members" : [
{
"_id" : 0,
"name" : "localhost:20001",
"health" : 1,
"state" : 1,
"stateStr" : "PRIMARY",
"optime" : {
"t" : 1321436824000,
"i" : 1
},
"optimeDate" : ISODate("2011-11-16T09:47:04Z"),
"self" : true
},
{
"_id" : 1,
"name" : "localhost:20002",
"health" : 1,
"state" : 2,
"stateStr" : "SECONDARY",
"uptime" : 203,
"optime" : {
"t" : 1321436824000,
"i" : 1
},
"optimeDate" : ISODate("2011-11-16T09:47:04Z"),
"lastHeartbeat" : ISODate("2011-11-16T09:50:28Z")
},
{
"_id" : 2,
"name" : "localhost:20003",
"health" : 1,
"state" : 2,
"stateStr" : "SECONDARY",
"uptime" : 203,
"optime" : {
"t" : 1321436824000,
"i" : 1
},
"optimeDate" : ISODate("2011-11-16T09:47:04Z"),
"lastHeartbeat" : ISODate("2011-11-16T09:50:28Z")
},
{
"_id" : 3,
"name" : "localhost:20005",
"health" : 1,
"state" : 2,
"stateStr" : "SECONDARY",
"uptime" : 185,
"optime" : {
"t" : 1321436824000,
"i" : 1
},
"optimeDate" : ISODate("2011-11-16T09:47:04Z"),
"lastHeartbeat" : ISODate("2011-11-16T09:50:28Z")
}
],
"ok" : 1
}

mongo 127.0.0.1:20005
MongoDB shell version: 1.8.4
connecting to: 127.0.0.1:20005/test
myrs:SECONDARY> show collections
Wed Nov 16 17:51:01 uncaught exception: error: { "$err" : "not master and slaveok=false", "code" : 13435 }
myrs:SECONDARY> db.getMongo().setSlaveOk()
not master and slaveok=false
myrs:SECONDARY> db.bank.find()
{ "_id" : 0, "name" : "joe", "age" : "50", "score" : 99 }
{ "_id" : 1, "name" : "joe", "age" : "60", "score" : 99 }
{ "_id" : 2, "name" : "joe", "age" : "70", "score" : 99 }
{ "_id" : ObjectId("4ec3849baa5bff10ca03145b"), "computer" : "mac" }

可以看到数据已经同步过来。

注：如果启动新节点或者kill -9杀掉进程重启的时候，有以下报错：

Wed Nov 16 17:45:06 [initandlisten] MongoDB starting :pid=15397port=20005 dbpath=/mongo/data5/ 64-bit
Wed Nov 16 17:45:06 [initandlisten] db version v1.8.4, pdfile version 4.5
Wed Nov 16 17:45:06 [initandlisten] git version: 81f12749a15e3d158b1b16bab6bc3faea538e166
Wed Nov 16 17:45:06 [initandlisten] build sys info: Linux bs-linux64.10gen.cc 2.6.21.7-2.ec2.v1.2.fc8xen #1 SMP Fri Nov 20 17:48:28 EST 2009 x86_64 BOOST_LIB
_VERSION=1_41
**************
old lock file: /mongo/data5/mongod.lock. probably means unclean shutdown
recommend removing file and running --repair
see:http://dochub.mongodb.org/core/repairfor more information
*************
Wed Nov 16 17:45:06 [initandlisten] exception in initAndListen std::exception: old lock file, terminating
Wed Nov 16 17:45:06 dbexit:
Wed Nov 16 17:45:06 [initandlisten] shutdown: going to close listening sockets...
Wed Nov 16 17:45:06 [initandlisten] shutdown: going to flush diaglog...
Wed Nov 16 17:45:06 [initandlisten] shutdown: going to close sockets...
Wed Nov 16 17:45:06 [initandlisten] shutdown: waiting for fs preallocator...
Wed Nov 16 17:45:06 [initandlisten] shutdown: closing all files...
Wed Nov 16 17:45:06 closeAllFiles() finished
Wed Nov 16 17:45:06 dbexit: really exiting now

这是由于启动新进程的时候去读取这个文件，这个文件其实就是一个进程号，以上红色标注的pid，那么就把数据目录下的mongod.lock删除掉即可，启动的时候会自动重建一个文件，重新分配一个pid。