MongoDB 3.2容灾测试

• 1台Primary,1台Secondary,1台Arbiter
  • 当Arbiter挂了
    • Primary和Secondary的主从不会切换
    • 外部应用服务不受影响

    • 在Primary上重新配置拓扑，去掉原有Aribter引入新的Arbiter, 命令如下。这个操作对Primary，Secondary以及外部应用都没有影响。

      rs.reconfig({"_id" : "rs1","members" : [{"_id" : 0, "host" : "127.0.0.1:27017","priority":1},{"_id" : 1, "host" : "127.0.0.1:27018","priority":2},{"_id" : 2, "host" : "127.0.0.1:27019","arbiterOnly" : true}]});

      
       

  • Primary和Arbiter同时挂了
    • Secondary一直保持Secondary状态，不会自动切换为Primary。重启之后，仍然保持Secondary状态。
    • 外部应用服务不能读写，且应用服务不能正常重启
    • 重启Primary，则经过几秒钟之后，Primary恢复成Primary状态，则外部应用能正常重启，外部应用读写正常，如果外部应用没有重启，则仍然能恢复正常读写；Primary与Secondary之间数据同步恢复正常。

    • 如果因为硬件等原因Primary不能重启，则需要在Secondary的上强制修改拓扑配置，将其余节点去掉，并将原来的Secondary指定为新的Primary，然后修改应用服务的mongo URL指向新的Primary再重启应用服务。修改命令如下：

      rs1:SECONDARY> use admin switched to db admin rs1:SECONDARY> db.auth('admin','adminpassword') 1 rs1:SECONDARY> rs.reconfig({"_id" : "rs1","members" : [{"_id" : 1, "host" : "127.0.0.1:27018","priority":2}]},{ force: true }); { "ok" : 1 } rs1:SECONDARY> rs.status() {         "set" : "rs1",         "date" : ISODate("2018-11-09T07:02:58.605Z"),         "myState" : 1,         "term" : NumberLong(24),         "heartbeatIntervalMillis" : NumberLong(2000),         "members" : [                 {                         "_id" : 1,                         "name" : "127.0.0.1:27018",                         "health" : 1,                         "state" : 1,                         "stateStr" : "PRIMARY",                         "uptime" : 1662,                         "optime" : {                                 "ts" : Timestamp(1541746976, 2),                                 "t" : NumberLong(24)                         },                         "optimeDate" : ISODate("2018-11-09T07:02:56Z"),                         "infoMessage" : "could not find member to sync from",                         "electionTime" : Timestamp(1541746976, 1),                         "electionDate" : ISODate("2018-11-09T07:02:56Z"),                         "configVersion" : 281693,                         "self" : true                 }         ],         "ok" : 1 } rs1:PRIMARY>

      等原有Primary恢复以后，再将其作为Secondary加入到拓扑当中，需要在新的Primary上修改配置，命令如下

      rs1:PRIMARY> rs.reconfig({"_id" : "rs1","members" : [{"_id" : 0, "host" : "127.0.0.1:27017","priority":1},{"_id" : 1, "host" : "127.0.0.1:27018","priority":2},{"_id" : 2, "host" : "127.0.0.1:27019","arbiterOnly" : true}]});

• 1台Primary,2台Secondary,1台Arbiter
  • Primary和Arbiter同时挂了，达到总节点数的一半
    • 2台Secondary无法自动选出一个Primary，日志中会报“Not starting an election, since we are not electable due to: Not standing for election because I cannot see a majority (mask 0x1)”
• 1台Primary,3台Secondary,1台Arbiter
  • Primary和Arbiter同时挂了
    • 3台Secondary会自动选出一个Primary,先选priority值最大的，priority值相同的情况则随机选择。
  • Primary和其中一个Secondary同时挂了
    • 2台Secondary中会自动选出一个Primary,先选priority值最大的，priority值相同的情况则随机选择。
  • Primary，其中一个Secondary和Arbiter同时挂了
    • 2台Secondary无法自动选出一个Primary，日志中会报“Not starting an election, since we are not electable due to: Not standing for election because I cannot see a majority (mask 0x1)”
• 1台Primary,4台Secondary,1台Arbiter
  • Primary和Arbiter同时挂了
    • 4台Secondary会自动选出一个Primary，暂时没有发现有平局的情况
  • Primary,Arbiter和一个Secondary同时挂，达到总节点数的一半
    • 3台Secondary无法自动选出一个Primary，日志中会报“Not starting an election, since we are not electable due to: Not standing for election because I cannot see a majority (mask 0x1)”
  • Primary和2个Secondary同时挂，达到总节点数的一半
    • 2台Secondary无法自动选出一个Primary，日志中会报“Not starting an election, since we are not electable due to: Not standing for election because I cannot see a majority (mask 0x1)”

多个Secondary节点并不是全部都从主节点同步，也可以从其它从节点同步。Secondaries may automatically change their sync from source as needed based on changes in the ping time and state of other members’ replication.

应用服务里所配的多个Mongo IP只要有一个还处于存活状态，则读写正常，应用服务重启也正常。