MongoDBcopy集tutorial

LearningMongoDBcopy集 原理, configurationmethod, failure转移 and monitor, 构建high availability MongoDBcluster

1. copy集原理

MongoDBcopy集 is 一组maintenance相同data集 MongoDBserver, providingdata冗余 and high availability性. copy集通常由一个主node (Primary) and many 个 from node (Secondary) 组成.

1.1 copy集architecture

copy集 basicarchitectureincluding:

  • 主node (Primary) : processing所 has 写operation, 记录operation to oplog in
  • from node (Secondary) : copy主node oplog 并application to 自己 data集
  • 仲裁node (Arbiter) : 不storedata, 仅参 and 选举过程

1.2 copymechanism

MongoDBusing oplog (operationlog) 来implementationcopy:

  • 主node将所 has 写operation记录 to oplog in
  • from node定期轮询主node获取 new oplog 条目
  • from nodeapplication这些operation to 自己 data集, 保持 and 主nodesynchronization
// 查看 oplog  big  small 
rs.printReplicationInfo();

2. configurationmethod

configurationMongoDBcopy集 步骤such as under :

2.1 准备server

至 few 需要3台server来构建一个basic copy集 (1个主node, 2个 from node) .

2.2 configurationMongoDBinstance

in 每台server on 启动MongoDBinstance, 并指定copy集名称: 

// server1
mongod --port 27017 --dbpath /data/db1 --replSet rs0

// server2
mongod --port 27017 --dbpath /data/db2 --replSet rs0

// server3
mongod --port 27017 --dbpath /data/db3 --replSet rs0

2.3 初始化copy集

连接 to 其in一个MongoDBinstance, 初始化copy集: 

// 连接 to MongoDB
mongo --port 27017

// 初始化copy集
rs.initiate({
   _id: "rs0",
   members: [
      { _id: 0, host: "server1:27017" },
      { _id: 1, host: "server2:27017" },
      { _id: 2, host: "server3:27017" }
   ]
});

2.4 verificationcopy集status

using rs.status() commandscheckcopy集status: 

// checkcopy集status
rs.status();

3. failure转移

MongoDBcopy集 failure转移mechanism确保 in 主node不可用时, systemable to自动选举 new 主node.

3.1 选举过程

当主node不可用时, from node会触发选举过程:

  • 检测 to 主nodefailure (心跳检测失败)
  • from node自我提名
  • othernode投票
  • 获得 many 数票 node成 for new 主node

3.2 选举条件

node被选 for new 主node 条件:

  • able to and big many 数node通信
  • datasynchronization程度 high (oplog 最 new )
  • priority (priority) 值 high (such as果设置)

3.3 failure转移时间

failure转移通常需要 10-30 秒completion, including:

  • 检测主nodefailure 时间 (默认10秒)
  • 选举 new 主node 时间 (通常几秒钟)

注意: in failure转移期间, copy集可能暂时无法processing写operation, 但仍可以processing读operation ( from from node) .

4. monitor

monitorMongoDBcopy集 healthystatus for 于确保system stable run至关 important .

4.1 in 置monitorcommands

// checkcopy集status
rs.status()

// checkcopylatency
rs.printSlaveReplicationInfo()

// checkcopyinformation
rs.printReplicationInfo()

4.2 关键monitor指标

  • copylatency: from node and 主node 时间差
  • oplog 窗口: oplog 覆盖 时间范围
  • 心跳status: node间通信status
  • 选举status: is 否 has 正 in for 选举

4.3 out 部monitortool

  • MongoDB Atlas: MongoDB官方 云monitorservice
  • MongoDB Compass: graph形化managementtool, package含monitorfunctions
  • Nagios/Zabbix: through插件monitorMongoDB
  • Prometheus + Grafana: usingMongoDB exporter收集指标

5. 实践case: 构建high availabilityMongoDBcluster

fake设我们需要 for 一个电商system构建high availability MongoDBcluster, 具体步骤such as under :

5.1 clusterplanning

  • 3台server: server1 (主node) , server2 ( from node) , server3 ( from node)
  • 端口: 27017
  • copy集名称: rs0

5.2 configuration步骤

// 1.  in 所 has server on creationdataTable of Contents
mkdir -p /data/db

// 2.  in 所 has server on 启动MongoDBinstance
mongod --port 27017 --dbpath /data/db --replSet rs0 --bind_ip_all

// 3. 连接 to 其in一个instance并初始化copy集
mongo --port 27017
rs.initiate({
   _id: "rs0",
   members: [
      { _id: 0, host: "server1:27017" },
      { _id: 1, host: "server2:27017" },
      { _id: 2, host: "server3:27017" }
   ]
});

// 4. verificationcopy集status
rs.status()

5.3 testfailure转移

// 1. 查看当 before 主node
rs.isMaster()

// 2. mock主nodefailure ( in 主node on 执行) 
use admin
db.shutdownServer()

// 3. check new  主node ( in  from node on 执行) 
rs.isMaster()

6. 互动练习

issues1: MongoDBcopy集 作用 is what？

A. improvingqueryperformance

B. providingdata冗余 and high availability性

C. 增加store容量

D. 简化management

issues2: copy集 最 small 推荐configuration is what？

A. 1个node

B. 2个node

C. 3个node

D. 5个node

issues3: failure转移过程in, copy集 is 否可以processing写operation？

A. 可以, 完全不影响

B. 暂时不可用, 直 to new 主node选举completion

C. 只能processing部分写operation

D. 取决于configuration

issues4: such as何查看copylatency？

A. rs.status()

B. rs.printSlaveReplicationInfo()

C. rs.printReplicationInfo()

D. db.serverStatus()

7. 推荐链接