mongodb 分片小计

之前以为启动mongod的时候配置好分片参数就直接分片了，还是太业余了，还需要两步：
1. 为db开启分片功能
2. 对开启分片功能的db的collection指定片键并分片
分别是这几个命令

# 启用数据库分片
db.runCommand({"enablesharding":"test"})
# 启用数据集分片
db.runCommand({"shardcollection" : "test.users", "key" : {"_id" : 1,"uid":1}})
# 在mongos shell查看分片状态
sh.status()

下面是一个例子

--- Sharding Status --- 
  sharding version: {
    "_id" : 1,
    "version" : 4,
    "minCompatibleVersion" : 4,
    "currentVersion" : 5,
    "clusterId" : ObjectId("578cbd23665b83cc995bedd1")
}
  shards:
    {  "_id" : "shard1",  "host" : "shard1/172.17.10.227:27021,172.17.10.228:27021,172.17.10.229:27021",  "maxSize" : NumberLong(20480) }
    {  "_id" : "shard2",  "host" : "shard2/172.17.10.227:27022,172.17.10.228:27022,172.17.10.229:27022",  "maxSize" : NumberLong(20480) }
  most recently active mongoses:
    "2.6.6" : 1
  balancer:
    Currently enabled:  yes
    Currently running:  yes
        Balancer lock taken at Sun Oct 09 2016 16:33:40 GMT+0800 (CST) by 172-17-10-227:30000:1474867375:1804289383:Balancer:846930886
    Collections with active migrations: 
        reducer-monitor.db_eth0_bps_in_metric started at Sun Oct 09 2016 16:57:19 GMT+0800 (CST)
  databases:
    {  "_id" : "admin",  "partitioned" : false,  "primary" : "config" }
    {  "_id" : "test",  "partitioned" : false,  "primary" : "shard2" }
    {  "_id" : "db",  "partitioned" : false,  "primary" : "shard1" }
    {  "_id" : "reducer-monitor",  "partitioned" : true,  "primary" : "shard1" }
        reducer-monitor.db_conntrack_rate_metric
            shard key: { "host" : 1, "BusinessMinute" : 1 }
            unique: false
            balancing: true
            chunks:
                shard1  12
                shard2  8
            too many chunks to print, use verbose if you want to force print
        reducer-monitor.db_cpu_metric
            shard key: { "host" : 1 }
            unique: false
            balancing: true
            chunks:
                shard1  194
                shard2  2
            too many chunks to print, use verbose if you want to force print
        reducer-monitor.db_eth0_bps_in_metric
            shard key: { "host" : 1, "BusinessMinute" : 1 }
            unique: false
            balancing: true
            chunks:
                shard1  193
            too many chunks to print, use verbose if you want to force print
        reducer-monitor.db_eth0_bps_out_metric
            shard key: { "host" : 1, "BusinessMinute" : 1 }
            unique: false
            balancing: true
            chunks:
                shard1  192
            too many chunks to print, use verbose if you want to force print
    {  "_id" : "amin",  "partitioned" : false,  "primary" : "shard2" }
    {  "_id" : "uhelm",  "partitioned" : false,  "primary" : "shard2" }
    {  "_id" : "uadb",  "partitioned" : false,  "primary" : "shard2" }

它正在banlancing

# 查看balancer的状态
sh.getBalancerState()
# 停止分片
sh.stopBalancer()
# 开始分片
sh.startBalancer()

true 表示正在执行，可以banlancer配置平衡的时间段

---

下一步我要将删除mongo数据造成的空洞数据释放，方法介绍如下：

把已有的空洞数据，remove掉，重新生成一份无空洞数据。那么具体如何落地？先预热从库；把预热的从库提升为主库；把之前主库的数据全部删除；重新同步；同步完成后，预热此库；把此库提升为主库。

具体操作如下：

检查服务器各节点是否正常运行 (ps -ef |grep mongod)；
登陆要处理的主节点 /mongodb/bin/mongo–port 88888；
做降权处理rs.stepDown()，并通过命令 rs.status()来查看是否降权；
切换成功之后，停掉该节点；检查是否已经降权，可以通过web页面查看status，我们建议最好登录进去保证有数据进入，或者是mongostat 查看； kill 掉对应mongo的进程： kill 进程号；删除数据，进入对应的分片删除数据文件，比如： rm -fr /mongodb/shard11/*；重新启动该节点，执行重启命令，比如：如:/mongodb/bin/mongod –config /mongodb/shard11.conf；通过日志查看进程；数据同步完成后，在修改后的主节点上执行命令 rs.stepDown() ，做降权处理。