MongoDB副本集的常用操作及原理

本文是对MongoDB副本集常用操作的一个汇总，同时也穿插着介绍了操作背后的原理及注意点。

结合之前的文章：MongoDB副本集的搭建，大家可以在较短的时间内熟悉MongoDB的搭建和管理。

下面的操作主要分为两个部分：

1. 修改节点状态

主要包括：

1> 将Primary节点降级为Secondary节点

2> 冻结Secondary节点

3> 强制Secondary节点进入维护模式

2. 修改副本集的配置

1> 添加节点

2> 删除节点

3> 将Secondary节点设置为延迟备份节点

4> 将Secondary节点设置为隐藏节点

5> 替换当前的副本集成员

6> 设置副本集节点的优先级

7> 阻止Secondary节点升级为Primary节点

8> 如何设置没有投票权的Secondary节点

9> 禁用chainingAllowed

10> 为Secondary节点显式指定复制源

11> 禁止Secondary节点创建索引

首先查看MongoDB副本集支持的所有操作

> rs.help()
    rs.status()                                { replSetGetStatus :  } checks repl set status
    rs.initiate()                              { replSetInitiate : null } initiates set with default settings
    rs.initiate(cfg)                           { replSetInitiate : cfg } initiates set with configuration cfg
    rs.conf()                                  get the current configuration object from local.system.replset
    rs.reconfig(cfg)                           updates the configuration of a running replica set with cfg (disconnects)
    rs.add(hostportstr)                        add a new member to the set with default attributes (disconnects)
    rs.add(membercfgobj)                       add a new member to the set with extra attributes (disconnects)
    rs.addArb(hostportstr)                     add a new member which is arbiterOnly:true (disconnects)
    rs.stepDown([stepdownSecs, catchUpSecs])   step down as primary (disconnects)
    rs.syncFrom(hostportstr)                   make a secondary sync from the given member
    rs.freeze(secs)                            make a node ineligible to become primary for the time specified
    rs.remove(hostportstr)                     remove a host from the replica set (disconnects)
    rs.slaveOk()                               allow queries on secondary nodes

    rs.printReplicationInfo()                  check oplog size and time range
    rs.printSlaveReplicationInfo()             check replica set members and replication lag
    db.isMaster()                              check who is primary

    reconfiguration helpers disconnect from the database so the shell will display
    an error, even if the command succeeds.

修改节点状态

将Primary节点降级为Secondary节点

myapp:PRIMARY> rs.stepDown()

这个命令会让primary降级为Secondary节点，并维持60s，如果这段时间内没有新的primary被选举出来，这个节点可以要求重新进行选举。

也可手动指定时间

myapp:PRIMARY> rs.stepDown()

在执行完该命令后，原Secondary node3:27017升级为Primary。

其日志输出为：

--03T22::21.009+ I COMMAND  [conn8] Attempting to step down in response to replSetStepDown command
--03T22::25.967+ I -        [conn8] end connection 127.0.0.1: ( connections now open)
--03T22::37.643+ I REPL     [ReplicationExecutor] Member node3: is now in state SECONDARY
--03T22::41.123+ I REPL     [replication-] Restarting oplog query due to error: InterruptedDueToReplStateChange: operat
ion was interrupted. Last fetched optime (with hash): { ts: Timestamp |, t:  }[-]. Restarts remaining: --03T22::41.167+ I REPL     [replication-] Scheduled new oplog query Fetcher source: node3: database: local query:
 { find: "oplog.rs", filter: { ts: { $gte: Timestamp | } }, tailable: true, oplogReplay: true, awaitData: true, maxTimeMS: , term:  } query metadata: { $replData: , $ssm: { $secondaryOk: true } } active:  timeout: 10000ms shutting down?:  first:  firstCommandScheduler: RemoteCommandRetryScheduler request: RemoteCommand  -- target:node3: db:local cmd:{ find: "oplog.rs", filter: { ts: { $gte: Timestamp | } }, tailable: true, oplogReplay: true, awaitData: true, maxTimeMS: , term:  } active:  callbackHandle.valid:  callbackHandle.cancelled:  attempt:  retryPolicy: RetryPolicyImpl maxAttempts:  maxTimeMillis: -1ms2017--03T22::41.265+ I REPL     [replication-] Choosing new sync source because our current sync source, node3:, has a
n OpTime ({ ts: Timestamp |, t:  }) which is not ahead of ours ({ ts: Timestamp |, t:  }), it does not have a sync source, and it's not the primary (sync source does not know the primary)2017-05-03T22:24:41.266+0800 I REPL     [replication-39] Canceling oplog query because we have to choose a new sync source. Current s
ource: node3:, OpTime { ts: Timestamp |, t: - }, its sync source index:---03T22::41.266+ W REPL     [rsBackgroundSync] Fetcher stopped querying remote oplog with error: InvalidSyncSource: sync
source node3: (last visible optime: { ts: Timestamp |, t: - }; config version: ; sync source index: -; primary index: -) is no longer valid2017--03T22::41.266+ I REPL     [rsBackgroundSync] could not find member to sync from
--03T22::46.021+ I REPL     [SyncSourceFeedback] SyncSourceFeedback error sending update to node3:: InvalidSyncSourc
e: Sync source was cleared. Was node3:--03T22::46.775+ I REPL     [ReplicationExecutor] Starting an election, since we've seen no PRIMARY in the past 10000ms
--03T22::46.775+ I REPL     [ReplicationExecutor] conducting a dry run election to see if we could be elected
--03T22::46.857+ I REPL     [ReplicationExecutor] VoteRequester(term  dry run) received a yes vote from node3:; res
ponse message: { term: , voteGranted: true, reason: "", ok: 1.0 }--03T22::46.858+ I REPL     [ReplicationExecutor] dry election run succeeded, running for election
--03T22::46.891+ I REPL     [ReplicationExecutor] VoteRequester(term ) received a yes vote from node3:; response me
ssage: { term: , voteGranted: true, reason: "", ok: 1.0 }--03T22::46.891+ I REPL     [ReplicationExecutor] election succeeded, assuming primary role in term
--03T22::46.891+ I REPL     [ReplicationExecutor] transition to PRIMARY
--03T22::46.892+ I ASIO     [NetworkInterfaceASIO-Replication-] Connecting to node3:
--03T22::46.894+ I ASIO     [NetworkInterfaceASIO-Replication-] Connecting to node3:
--03T22::46.894+ I ASIO     [NetworkInterfaceASIO-Replication-] Successfully connected to node3:
--03T22::46.895+ I REPL     [ReplicationExecutor] My optime is most up-to-date, skipping catch-up and completing transiti
on to primary.--03T22::46.895+ I ASIO     [NetworkInterfaceASIO-Replication-] Successfully connected to node3:
--03T22::47.348+ I REPL     [rsSync] transition to primary complete; database writes are now permitted
--03T22::49.231+ I NETWORK  [thread1] connection accepted from 192.168.244.30: # ( connections now open)
--03T22::49.236+ I NETWORK  [conn9] received client metadata from 192.168.244.30: conn9: { driver: { name: "NetworkI
nterfaceASIO-RS", version: "3.4." }, os: { type: "Linux", name: "Red Hat Enterprise Linux Server release 6.7 (Santiago)", architecture: "x86_64", version: "Kernel 2.6.-.el6.x86_64" } }2017-05-03T22:24:49.317+0800 I NETWORK  [thread1] connection accepted from 192.168.244.30:35838 #10 (4 connections now open)
--03T22::49.318+ I NETWORK  [conn10] received client metadata from 192.168.244.30: conn10: { driver: { name: "Networ
kInterfaceASIO-RS", version: "3.4." }, os: { type: "Linux", name: "Red Hat Enterprise Linux Server release 6.7 (Santiago)", architecture: "x86_64", version: "Kernel 2.6.-.el6.x86_64" } }

原Primary node3:27018降低为Secondary

--03T22::36.262+ I COMMAND  [conn7] Attempting to step down in response to replSetStepDown command
--03T22::36.303+ I REPL     [conn7] transition to SECONDARY
--03T22::36.315+ I NETWORK  [conn7] legacy transport layer closing all connections
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.316+ I NETWORK  [conn7] Skip closing connection for connection #
--03T22::36.382+ I NETWORK  [thread1] connection accepted from 127.0.0.1: # ( connections now open)
--03T22::36.383+ I NETWORK  [conn8] received client metadata from 127.0.0.1: conn8: { application: { name: "MongoDB
Shell" }, driver: { name: "MongoDB Internal Client", version: "3.4." }, os: { type: "Linux", name: "Red Hat Enterprise Linux Server release 6.7 (Santiago)", architecture: "x86_64", version: "Kernel 2.6.-.el6.x86_64" } }2017-05-03T22:24:36.408+0800 I -        [conn7] AssertionException handling request, closing client connection: 172 Operation attempt
ed on a closed transport Session.--03T22::36.408+ I -        [conn7] end connection 127.0.0.1: ( connections now open)
--03T22::41.262+ I COMMAND  [conn5] command local.oplog.rs command: find { find: "oplog.rs", filter: { ts: { $gte: Timest
amp | } }, tailable: true, oplogReplay: true, awaitData: true, maxTimeMS: , term:  } planSummary: COLLSCAN cursorid: keysExamined: docsExamined: writeConflicts: numYields: nreturned: reslen: locks:{ Global: { acquireCount: { r:  } }, Database: { acquireCount: { r:  } }, oplog: { acquireCount: { r:  } } } protocol:op_command 100ms2017--03T22::48.311+ I REPL     [ReplicationExecutor] Member node3: is now in state PRIMARY
--03T22::49.163+ I REPL     [rsBackgroundSync] sync source candidate: node3:
--03T22::49.164+ I ASIO     [NetworkInterfaceASIO-RS-] Connecting to node3:
--03T22::49.236+ I ASIO     [NetworkInterfaceASIO-RS-] Successfully connected to node3:
--03T22::49.316+ I ASIO     [NetworkInterfaceASIO-RS-] Connecting to node3:
--03T22::49.318+ I ASIO     [NetworkInterfaceASIO-RS-] Successfully connected to node3:
--03T22::41.020+ I -        [conn4] end connection 192.168.244.30: ( connections now open)
--03T22::02.653+ I ASIO     [NetworkInterfaceASIO-RS-] Connecting to node3:
--03T22::02.669+ I ASIO     [NetworkInterfaceASIO-RS-] Successfully connected to node3:
--03T22::41.442+ I -        [conn5] end connection 192.168.244.30: ( connections now open)

冻结Secondary节点

如果需要对Primary做一下维护，但是不希望在维护的这段时间内将其它Secondary节点选举为Primary节点，可以在每次Secondary节点上执行freeze命令，强制使它们始终处于Secondary节点状态。

myapp:SECONDARY> rs.freeze()

注：只能在Secondary节点上执行

myapp:PRIMARY> rs.freeze()
{
    "ok" : ,
    "errmsg" : "cannot freeze node when primary or running for election. state: Primary",
    "code" : ,
    "codeName" : "NotSecondary"
}

如果要解冻Secondary节点，只需执行

myapp:SECONDARY> rs.freeze()

强制Secondary节点进入维护模式

当Secondary节点进入到维护模式后，它的状态即转化为“RECOVERING”，在这个状态的节点，客户端不会发送读请求给它，同时它也不能作为复制源。

进入维护模式有两种触发方式：

1. 自动触发

譬如Secondary上执行压缩

2. 手动触发

myapp:SECONDARY> db.adminCommand({"replSetMaintenance":true})

修改副本集的配置

添加节点

myapp:PRIMARY> rs.add("node3:27017")

myapp:PRIMARY> rs.add({_id: , host: "node3:27017", priority: , hidden: true})

也可通过配置文件的方式

> cfg={
    "_id" : ,
    "host" : "node3:27017",
    "arbiterOnly" : false,
    "buildIndexes" : true,
    "hidden" : true,
    "priority" : ,
    "tags" : {

    },
    "slaveDelay" : NumberLong(),
    "votes" :
}
> rs.add(cfg)

删除节点

第一种方式

myapp:PRIMARY> rs.remove("node3:27017")

第二种方式

myapp:PRIMARY> cfg = rs.conf()
myapp:PRIMARY> cfg.members.splice(,)
myapp:PRIMARY> rs.reconfig(cfg)

注：执行rs.reconfig并不必然带来副本集的重新选举，加force参数同样如此。

The rs.reconfig() shell method can trigger the current primary to step down in some situations. 

修改节点的配置

将Secondary节点设置为延迟备份节点

cfg = rs.conf()
cfg.members[].priority =
cfg.members[].hidden = true
cfg.members[].slaveDelay =
rs.reconfig(cfg)

将Secondary节点设置为隐藏节点

cfg = rs.conf()
cfg.members[].priority =
cfg.members[].hidden = true
rs.reconfig(cfg)

替换当前的副本集成员

cfg = rs.conf()
cfg.members[].host = "mongo2.example.net"
rs.reconfig(cfg)

设置副本集节点的优先级

cfg = rs.conf()
cfg.members[].priority = 0.5
cfg.members[].priority =
cfg.members[].priority =
rs.reconfig(cfg)

优先级的有效取值是0~1000，可为小数，默认为1

从MongoDB 3.2开始

Non-voting members must have priority of .
Members with priority greater than  cannot have  votes.

注：如果将当前Secondary节点的优先级设置的大于Primary节点的优先级，会导致当前Primary节点的退位。

阻止Secondary节点升级为Primary节点

只需将priority设置为0

fg = rs.conf()
cfg.members[].priority =
rs.reconfig(cfg)

如何设置没有投票权的Secondary节点

MongoDB限制一个副本集最多只能拥有50个成员节点，其中，最多只有7个成员节点拥有投票权。

之所以作此限制，主要是考虑到心跳请求导致的网络流量，毕竟每个成员都要向其它所有成员发送心跳请求，和选举花费的时间。

从MongoDB 3.2开始，任何priority大于0的节点都不可将votes设置为0

所以，对于没有投票权的Secondary节点，votes和priority必须同时设置为0

cfg = rs.conf()
cfg.members[].votes =
cfg.members[].priority =
cfg.members[].votes =
cfg.members[].priority =
rs.reconfig(cfg) 

禁用chainingAllowed

默认情况下，允许级联复制。

即备份集中如果新添加了一个节点，这个节点很可能是从其中一个Secondary节点处进行复制，而不是从Primary节点处复制。

MongoDB根据ping时间选择同步源，一个节点向另一个节点发送心跳请求，就可以得知心跳请求所耗费的时间。MongoDB维护着不同节点间心跳请求的平均花费时间，选择同步源时，会选择一个离自己比较近而且数据比自己新的节点。

如何判断节点是从哪个节点处进行复制的呢？

myapp:PRIMARY> rs.status().members[].syncingTo
node3:

当然，级联复制也有显而易见的缺点：复制链越长，将写操作复制到所有Secondary节点所花费的时间就越长。

可通过如下方式禁用

cfg=rs.conf()
cfg.settings.chainingAllowed=false
rs.reconfig(cfg)

将chainingAllowed设置为false后，所有Secondary节点都会从Primary节点复制数据。

为Secondary节点显式指定复制源

rs.syncFrom("node3:27019")

禁止Secondary节点创建索引

有时，并不需要Secondary节点拥有和Primary节点相同的索引，譬如这个节点只是用来处理数据备份或者离线的批量任务。这个时候，就可以阻止Secondary节点创建索引。

在MongoDB 3.4版本中，不允许直接修改，只能在添加节点时显式指定

myapp:PRIMARY> cfg=rs.conf()
myapp:PRIMARY> cfg.members[].buildIndexes=false
false
myapp:PRIMARY> rs.reconfig(cfg)
{
    "ok" : ,
    "errmsg" : "priority must be 0 when buildIndexes=false",
    "code" : ,
    "codeName" : "NewReplicaSetConfigurationIncompatible"
}
myapp:PRIMARY> cfg.members[].buildIndexes=false
false
myapp:PRIMARY> cfg.members[].priority=

myapp:PRIMARY> rs.reconfig(cfg)
{
    "ok" : ,
    "errmsg" : "New and old configurations differ in the setting of the buildIndexes field for member node3:27017; to make this c
hange, remove then re-add the member",    "code" : 103,
    "codeName" : "NewReplicaSetConfigurationIncompatible"
}
myapp:PRIMARY> rs.remove("node3:27017")
{ "ok" :  }
myapp:PRIMARY> rs.add({_id: , host: "node3:27017", priority: , buildIndexes:false})
{ "ok" :  }

从上述测试中可以看出，如果要将节点的buildIndexes设置为false，必须同时将priority设置为0。

参考

1.《MongoDB权威指南》

2. MongoDB官方文档