说明

公司现有线上mongo环境：

1. 不同版本（2.2.0、2.2.2、2.4.8、tokumx-2.0.0）

2. 不同架构环境（单实例、带master参数的单实例，分片+副本集）

为了将现有复杂的线上环境在不影响业务正常运行的条件下，迁移到同一套副本集环境中，下文模拟了不同版本、不同架构的情况下，来进行数据迁移到新的副本集（具体源环境见下节的源端环境）的实验

难点：由于源端是线上环境（随时数据会变化），在不影响业务正常使用（尽量不停机或者停机时间很短）的情况下迁移数据到新环境，同时要保证数据一致性、完整性

源端环境

公司目前现有mongo环境，如下

业务	主机	版本	环境	分类
--	--	2.2.2	单--master	2
--	--	2.2.0	单实例	1
--	--	2.2.2	分片+副本集	4
--	--	2.4.8	单--master	3
--	--	tokumx-2.0.0	单	5
--	--	2.2.0	单实例	1
--	--	2.2.0	单实例	1

#如上环境分为5类，下面将模拟出相同版本和相同环境的mongodb实例，进行迁移实验

目标环境

#3个node的副本集

yougou:PRIMARY> rs.status()

{

"set" : "yougou",

"date" : ISODate("2015-04-01T17:20:06Z"),

"myState" : 1,

"members" : [

{

"_id" : 0,

"name" : "192.168.211.250:27017",

"health" : 1,

"state" : 1,

"stateStr" : "PRIMARY",

"uptime" : 1061,

"optime" : Timestamp(1427908775, 1),

"optimeDate" : ISODate("2015-04-01T17:19:35Z"),

"self" : true

},

{

"_id" : 1,

"name" : "192.168.211.250:20000",

"health" : 1,

"state" : 2,

"stateStr" : "SECONDARY",

"uptime" : 34,

"optime" : Timestamp(1427908775, 1),

"optimeDate" : ISODate("2015-04-01T17:19:35Z"),

"lastHeartbeat" : ISODate("2015-04-01T17:20:04Z"),

"lastHeartbeatRecv" : ISODate("2015-04-01T17:20:05Z"),

"pingMs" : 0,

"syncingTo" : "192.168.211.250:27017"

},

{

"_id" : 2,

"name" : "192.168.211.250:30000",

"health" : 1,

"state" : 2,

"stateStr" : "SECONDARY",

"uptime" : 31,

"optime" : Timestamp(1427908775, 1),

"optimeDate" : ISODate("2015-04-01T17:19:35Z"),

"lastHeartbeat" : ISODate("2015-04-01T17:20:05Z"),

"lastHeartbeatRecv" : ISODate("2015-04-01T17:20:05Z"),

"pingMs" : 0,

"syncingTo" : "192.168.211.250:27017"

}

],

"ok" : 1

}

迁移工具mongosync

mongosync 是 mongodb 数据同步工具，目前已支持绝大多种数据同步。

注意：

1. mongosync通过mongo的oplog来实现增量同步，故如果要实现实时同步，则源端必须打开oplog功能，而oplog只有在mong复制、副本集环境下才会打开。于是，对于没有指定—master参数的单实例，需要重启mongdb，指定—master选项来实现

2. mongosync使用需要指定用户密码，故源端如果没有存在admin用户，需要addUser

实验内容

1.mongo-2.2.0单实例无master选项迁移到副本集

非实时同步-直接迁移

源端环境

/data/mongod/mongodb-2.2.0/bin/mongod --dbpath /data/mongodb/node1 --logpath /data/mongodb/logs/mongodb-2.2.0.log --logappend --fork --auth --port 22000 –directoryperdb

如上，启动选项看出无 master选项

~]# /data/mongod/mongodb-2.2.0/bin/mongo --port 22000

MongoDB shell version: 2.2.0

connecting to: 127.0.0.1:22000/test

> show dbs

admin (empty)

db220 0.203125GB

local (empty) #说明没有oplog

> use db220

switched to db db220

> show collections

system.indexes

t220

> db.t220.count()

888

目标端同步数据

~]# mongosync -h 192.168.211.217:22000 -u root -p yougou –d db220 --to 192.168.211.250:27017 -tu root -tp yougou --oplog

connected to: 192.168.211.217:22000

Thu Apr 2 01:57:07.632 [mongosync] 192.168.211.217:22000 connected ok

Thu Apr 2 01:57:07.634 [mongosync] admin auth ok

Thu Apr 2 01:57:07.635 [mongosync] 192.168.211.250:27017 connected ok

Thu Apr 2 01:57:07.636 [mongosync] admin auth ok

Thu Apr 2 01:57:07.639 [mongosync] lastOp OpTime:0,0 (Jan 1 08:00:00 0:0)

Thu Apr 2 01:57:07.639 [mongosync] clone all dbs ^_^

Thu Apr 2 01:57:07.641 [mongosync] DATABASE: db220 to db220

Thu Apr 2 01:57:07.910 [mongosync] cloning db220.t220 -> db220.t220

Thu Apr 2 01:57:07.927 [mongosync] 888 objects

Thu Apr 2 01:57:07.928 [mongosync] cloning db220.system.users -> db220.system.users

Thu Apr 2 01:57:07.930 [mongosync] 0 objects

Thu Apr 2 01:57:07.930 [mongosync] cloning db220.system.indexes -> db220.system.indexes

Thu Apr 2 01:57:07.932 [mongosync] 1 objects

Thu Apr 2 01:57:07.937 [mongosync] 0 objects

Thu Apr 2 01:57:07.940 [mongosync] clone done,start to catch up

Thu Apr 2 01:57:07.941 [mongosync] No new ops,waiting...

Thu Apr 2 01:57:12.942 [mongosync] No new ops,waiting...

……….

登录目标端验证

yougou:PRIMARY> db.t220.count()

888 #发现源端的db220同步完成

#源端修改数据（insert）

> for(var i=1;i<=888;i++) db.t220.save({"id":i,"a":123456789,"b":888888888,"c":100000000})

> db.t220.count()

1776

#目标端查看，新修改的数据没有同步

yougou:PRIMARY> db.t220.count()

888

故不能做到实时同步，为了解决此问题，则源端必须打开oplog，我们借助最简单的方式指定—master选项启动

实时数据同步—master

1.重启源端mongo实例

~]# netstat -ntpl|grep mongo|grep 22000

tcp 0 0 0.0.0.0:22000 0.0.0.0:* LISTEN 20696/mongod

]# kill 20696

--指定maste选项

]# /data/mongod/mongodb-2.2.0/bin/mongod --dbpath /data/mongodb/node1 --logpath /data/mongodb/logs/mongodb-2.2.0.log --logappend --fork --auth --port 22000 --directoryperdb –master

forked process: 5558

all output going to: /data/mongodb/logs/mongodb-2.2.0.log

child process started successfully, parent exiting

#登录源端，发现oplog已经打开

> use local

switched to db local

> show collections

oplog.$main

2.目标端执行同步命令

~]# mongosync -h 192.168.211.217:22000 -u root -p yougou -d db220 --to 192.168.211.250:27017 -tu root -tp yougou –oplog

Thu Apr 2 02:12:39.672 [mongosync] 192.168.211.217:22000 connected ok

Thu Apr 2 02:12:39.675 [mongosync] admin auth ok

Thu Apr 2 02:12:39.675 [mongosync] 192.168.211.250:27017 connected ok

Thu Apr 2 02:12:39.677 [mongosync] admin auth ok

Thu Apr 2 02:12:39.681 [mongosync] lastOp OpTime:1427941057,1 (Apr 2 10:17:37 551ca6c1:1)

Thu Apr 2 02:12:39.681 [mongosync] DATABASE: db220 to db220

Thu Apr 2 02:12:40.005 [mongosync] cloning db220.t220 -> db220.t220

Thu Apr 2 02:12:40.098 [mongosync] 1776 objects

Thu Apr 2 02:12:40.098 [mongosync] cloning db220.system.users -> db220.system.users

Thu Apr 2 02:12:40.102 [mongosync] 0 objects

Thu Apr 2 02:12:40.102 [mongosync] cloning db220.system.indexes -> db220.system.indexes

Thu Apr 2 02:12:40.106 [mongosync] 1 objects

Thu Apr 2 02:12:40.106 [mongosync] sync database: db220

Thu Apr 2 02:12:40.108 [mongosync] begin to apply oplog...

Thu Apr 2 02:12:40.108 [mongosync] db220:0 rows oplog to apply...

。。。。。。。。。。

#此时在源端新增加数据

> for(var i=1;i<=8888;i++) db.t220.save({"id":i,"a":123456789,"b":888888888,"c":100000000})

> db.t220.count()

10664

#目标端同步的日志可以发现，增量同步了

Thu Apr 2 02:13:06.158 [mongosync] No new oplogs,the data is the newest ^_^.waiting.

Thu Apr 2 02:13:07.175 [mongosync] Synced up to optime:ts: Timestamp 1427941089000|1

Thu Apr 2 02:13:12.650 [mongosync] 8888 ops

Thu Apr 2 02:13:12.650 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:16.403 [mongosync] 8888 ops

Thu Apr 2 02:13:16.403 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:21.415 [mongosync] 8888 ops

Thu Apr 2 02:13:21.415 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:25.420 [mongosync] 8888 ops

Thu Apr 2 02:13:25.420 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:30.432 [mongosync] 8888 ops

Thu Apr 2 02:13:30.432 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:34.435 [mongosync] 8888 ops

Thu Apr 2 02:13:34.435 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:39.447 [mongosync] 8888 ops

Thu Apr 2 02:13:39.448 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:44.297 [mongosync] 8888 ops

Thu Apr 2 02:13:44.297 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

Thu Apr 2 02:13:49.309 [mongosync] 8888 ops

Thu Apr 2 02:13:49.309 [mongosync] db220:8888 rows oplog are applied, waiting for new data ^_^.The latest optime is ts: Timestamp 1427941089000|8888

#同时登录目标端，可以看到表的记录在增加

yougou:PRIMARY> db.t220.count()

1776

yougou:PRIMARY> db.t220.count()

10664

同样在源端新创建集合也能实时同步，从而做到真正意义的同步

3.主要操作步骤

1. 重启源端单实例mongodb，启动的时候指定—master参数

2. 在目标端使用mongosync对-d xxx指定的数据库进行同步，指定—oplog可以做到实时同步

3. 同步跟上后，源端的上层应用可以切换到目标端

4. 待目标端使用确定没问题后，源端可以销毁

2.mongo-2.2.2并且有master选项迁移到副本集

直接进行实时同步，同上面1的实时数据同步—master

~]# mongosync -h 192.168.211.217:22200 -u root -p yougou -d db222 --to 192.168.211.250:27017 -tu root -tp yougou --oplog

3.mongo-2.4.8并且有master选项迁移到副本集

同上

~]# mongosync -h 192.168.211.217:24800 -u root -p yougou -d db248 --to 192.168.211.250:27017 -tu root -tp yougou –oplog

4.分片（副本集）同步到副本集

~]# mongosync -h 192.168.211.217:20000 -d yougou --to 192.168.211.250:27017 -tu root -tp yougou --oplog

connected to: 192.168.211.217:20000

Thu Apr 2 05:29:34.491 [mongosync] 192.168.211.217:20000 connected ok

Thu Apr 2 05:29:34.492 [mongosync] 192.168.211.250:27017 connected ok

Thu Apr 2 05:29:34.549 [mongosync] admin auth ok

Thu Apr 2 05:29:34.554 [mongosync] stop balancer

Thu Apr 2 05:29:34.557 [mongosync] target shards are less than source shards.

#由于mongosync暂时不支持分片同步到副本集，需要单独开多个mongosync分别连到各个分片来同步，有几个分片开几个mongosync,并且如果要实现实时同步，就必须用到oplog增量，所以源端的分片必须副本集或者主从（总之oplog开启）

~]# /data/mongod/mongodb-2.2.2/bin/mongo --port 20000

MongoDB shell version: 2.2.2

connecting to: 127.0.0.1:20000/test

mongos> use config

switched to db config

mongos> db.shards.find()

{ "_id" : "shard1", "host" : "shard1/192.168.211.217:30000,192.168.211.217:40000" }

{ "_id" : "shard2", "host" : "shard2/192.168.211.217:50000,192.168.211.217:60000" }

#如上，有2个分片shard1和shard2分别为2节点的副本集

源端模拟实时写入数据

mongos>for(var i=1;i<=888888;i++) db.yujx.save({"id":i,"a":123456789,"b":888888888,"c":100000000})

mongos>for(var i=1;i<=888888;i++) db.yujx.save({"id":0,"a":123456789,"b":888888888,"c":100000000})

目标端开启mongosync同步

#同步分片shard1的

~]# mongosync -h 192.168.211.217:30000 -d yougou --to 192.168.211.250:27017 -tu root -tp yougou –oplog

#同步分片shard2的

~]# mongosync -h 192.168.211.217:50000 -d yougou --to 192.168.211.250:27017 -tu root -tp yougou –oplog

验证实时同步

#最终，源端查询yujx集合状态可知：shard1上28929 row,shard2上357412 row,共386341行

mongos> db.yujx.stats()

{

"sharded" : true,

"ns" : "yougou.yujx",

"count" : 386341,

"numExtents" : 13,

"size" : 26271188,

"storageSize" : 40591360,

"totalIndexSize" : 26326720,

"indexSizes" : {

"_id_" : 12558336,

"id_1" : 13768384

},

"avgObjSize" : 68,

"nindexes" : 2,

"nchunks" : 3,

"shards" : {

"shard1" : {

"ns" : "yougou.yujx",

"count" : 28929,

"size" : 1967172,

"avgObjSize" : 68,

"storageSize" : 2793472,

"numExtents" : 5,

"nindexes" : 2,

"lastExtentSize" : 2097152,

"paddingFactor" : 1,

"systemFlags" : 1,

"userFlags" : 0,

"totalIndexSize" : 1766016,

"indexSizes" : {

"_id_" : 948416,

"id_1" : 817600

},

"ok" : 1

},

"shard2" : {

"ns" : "yougou.yujx",

"count" : 357412,

"size" : 24304016,

"avgObjSize" : 68,

"storageSize" : 37797888,

"numExtents" : 8,

"indexSizes" : {

"_id_" : 11609920,

"id_1" : 12950784

},

"ok" : 1

}

},

"ok" : 1

}

mongos> db.yujx.count()

386341

#到目标端查询集合yujx的情况

yougou:PRIMARY> db.yujx.count()

386341

yougou:PRIMARY> db.yujx.stats()

{

"ns" : "yougou.yujx",

"count" : 386341,

"size" : 27816584,

"avgObjSize" : 72.00008282838218,

"storageSize" : 37797888,

"systemFlags" : 1,

"userFlags" : 0,

"totalIndexSize" : 26948096,

"indexSizes" : {

"_id_" : 12541984,

"id_1" : 14406112

},

"ok" : 1

}#如上源端和目标端的数据室同步的

并且同步过程中mongosync通过日志，可以看到实时增量部分的情况，或者，在目标端隔时来运行db.yujx.count()可以看到，数据量一致增加的，从而做到了实时同步

5.tokumx-2.0.0同步到mongo副本集

mongosync的限制上写了，不适用于TokuMX同步数据到Mongo

故，此种情况为了保证数据完整性，必须给停机或者应用停止修改数据的时间窗口，在此期间导出数据并且导入mongo完成后，应用直接连新的mongo环境

实验结果

将如下源环境迁移到目标（副本集）环境

ITPub博客

Mongo数据迁移实验

说明

源端环境

目标环境

迁移工具mongosync

实验内容

1.mongo-2.2.0单实例无master选项迁移到副本集

非实时同步-直接迁移

源端环境

目标端同步数据

登录目标端验证

实时数据同步—master

1.重启源端mongo实例

2.目标端执行同步命令

3.主要操作步骤

2.mongo-2.2.2并且有master选项迁移到副本集

3.mongo-2.4.8并且有master选项迁移到副本集

4.分片（副本集）同步到副本集

源端模拟实时写入数据

目标端开启mongosync同步

验证实时同步

5.tokumx-2.0.0同步到mongo副本集

实验结果

最新文章