author:cjw 2019-6-13
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
规则范围分片算法支持=,between,in进行分区剪裁
配置样例
{"customTables":{},"globalTables":{},"normalTables":{},"schemaName":"db1","shardingTables":{"sharding":{"createTableSQL":"CREATE TABLE db1.`sharding` (\n `id` bigint NOT NULL AUTO_INCREMENT,\n `user_id` varchar(100) DEFAULT NULL,\n `create_time` date DEFAULT NULL,\n `fee` decimal(10,0) DEFAULT NULL,\n `days` int DEFAULT NULL,\n `blob` longblob,\n PRIMARY KEY (`id`),\n KEY `id` (`id`)\n) ENGINE=InnoDB DEFAULT CHARSET=utf8","function":{"clazz":"io.mycat.router.mycat1xfunction.XXXXXXX","properties":{"lastTime":90,"partionDay":180,"dateFormat":"yyyy-MM-dd","columnName":"create_time"},"ranges":{}},"partition":{"schemaNames":"db1","tableNames":"sharding_$0-1","targetNames":"c0"}}}}
在mycat2能正常启动的情况下,根据当前配置自动创建分片表,全局表的物理表(但不自动创建单表的物理表),如果物理表不存在
//在mycat终端输入/*+ mycat:repairPhysicalTable{} */;//1.20-2020-7-17后,实验
分片算法配置
defaultNode
一般来说,分片算法都允许配置defaultNode(默认节点),有部分没有实现不支持
当此值为-1时候,表示路由到默认节点报错
0代表第一个节点
ranges
分片键值的映射范围,多数情况下,值就是dataNode的下标,0代表第一个节点
键有单值和范围值的两种配置类型,
- 1000 : 0 一对一关系
- 1-2:0 一个连续范围对应一个dataNode1下标
具体看分片算法配置
以下内容引用mycat权威指南的一部分内容
properties
里面除了写分片算法属性外,必须写columnName设置分片字段
非分片
不计算分片值,指定dataNode的下标
function: { clazz: io.mycat.router.mycat1xfunction.PartitionConstant , properties: {defaultNode: '0',columnName: 'id'}} #映射到第一个dataNode
{"function": {"clazz": "io.mycat.router.mycat1xfunction.PartitionConstant","properties": {"defaultNode": "0","columnName": "id"}}}
分片枚举
通过在配置文件中配置可能的枚举 id，自己配置分片，本规则适用于特定的场景，比如有些业务需要按照省
份或区县来做保存，而全国省份区县固定的，这类业务使用本条规则，配置如下：
- clazz: io.mycat.router.mycat1xfunction.PartitionByFileMapname: PartitionByFileMapproperties:type: IntegerdefaultNode: -1columnName: 'id'ranges:10000: 010010: 1
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByFileMap","name": "PartitionByFileMap","properties": {"type": "Integer","defaultNode": -1,"columnName": "id"},"ranges": {"10000": 0,"10010": 1}}
/*+ mycat:createTable{"schemaName":"1cloud","shardingTable":{"createTableSQL":"create table sharding(\nid int(11) NOT NULL AUTO_INCREMENT,\nuser_id int(11) ,\nuser_name varchar(128), \nPRIMARY KEY (`id`) \n)ENGINE=InnoDB DEFAULT CHARSET=utf8 ","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByFileMap","properties":{"defaultNode":"0","type":"Integer","columnName":"id"},"ranges":{"130100":"0","130200":"1","130300":"2"}},"partition":{"schemaNames":"c","tableNames":"file_$0-2","targetNames":"prototype"}},"tableName":"sharding"} */;
使用本地文件配置范围(2021-7-20)
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByFileMap","name": "PartitionByFileMap","properties": {"type": "Integer","defaultNode": -1,"columnName": "id","mapFile":"partition-hash-int.txt"}}
启动的时候会提示partition-hash-int.txt放置的目录(2021-7-28)
PartitionByFileMap mapFile default path:
另外也支持绝对路径配置
partition-hash-int.txt
10000=010010=1
type
"type":"Integer",
分片字段使用的计算值的类型
Integer,Byte,Char,String,Long,Double,Float,Short,Boolean,BigInteger,BigDecimal
defaultNode
ranges
固定hash分片
本条规则类似于十进制的求模运算，区别在于是二进制的操作,是取 id 的二进制低 10 位，即 id 二进制
&1111111111。
此算法的优点在于如果按照 10 进制取模运算，在连续插入 1-10 时候 1-10 会被分到 1-10 个分片，增
大了插入的事务控制难度，而此算法根据二进制则可能会分到连续的分片，减少插入事务事务控制难度。
- clazz: io.mycat.router.mycat1xfunction.PartitionByLongname: partitionByLongproperties:partitionCount: 2,1partitionLength: 256,512columnName: 'id'
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByLong","name": "partitionByLong","properties": {"partitionCount": "2,1","partitionLength": "256,512","columnName": "id"}}
partitionCount
分片长度是分片范围的一种度量,在此分片算法中,总量是1024
因为分片总长度总量是1024
而且每个dataNode可以占有一个范围的分片长度
所以所有dataNode的分片长度和必须是1024
partitionCount配置一个数组
每个元素是分片节点的数量,所有元素之和就是分片节点数量
每个元素可以对应一个分片长度,在此元素上的每个分片节点使用该长度依次占有分片长度
partitionLength配置一个数组
每个元素是分片长度
partitionCount和partitionLength两个数组长度必须一致
上述例子中
1024 = 2X256+1X512
本例的分区策略：希望将数据水平分成 3 份，前两份各占 25%，第三份占 50%。（故本例非均匀分区）
|<———————1024———————————>||<—-256—>|<—-256—>|<———-512————->|| partition0 | partition1 | partition2 |
范围约定
- clazz: io.mycat.router.mycat1xfunction.AutoPartitionByLongname: AutoPartitionByLongproperties:defaultNode: -1columnName: 'id'ranges:0-500M: 0500M-1000M: 1
{"clazz": "io.mycat.router.mycat1xfunction.AutoPartitionByLong","name": "AutoPartitionByLong","properties": {"defaultNode": -1,"columnName": "id"},"ranges": {"0-500M": 0,"500M-1000M": 1}}
ranges填写多个范围值的映射,可以填写单位K,M
1K=1000
1M=10000
defaultNode 超过范围后的默认节点。
所有的节点配置都是从 0 开始，及 0 代表节点 1，此配置非常简单，即预先制定可能的 id 范围到某个分片
或
0-10000000=010000001-20000000=1
或
/*+ mycat:createTable{"schemaName":"db1","shardingTable":{"createTableSQL":"create table address3(\nid int(11) NOT NULL AUTO_INCREMENT,\nuser_id int(11) ,\nuser_name varchar(128), \nPRIMARY KEY (id) \n)ENGINE=InnoDB DEFAULT CHARSET=utf8 ","function":{"clazz":"io.mycat.router.mycat1xfunction.AutoPartitionByLong","properties": {"type":"Long","defaultNode": 1,"columnName": "id",},"ranges": {"0-1000":0,"1001-2000":1,"2001-3000":2}},"partition":{"schemaNames":"db3","tableNames":"addr3_$0-2","targetNames":"prototype"}},"tableName":"address3"} */;
2022-1-6后支持在properties写上mapFile属性指定ranges范围
取模
- clazz: io.mycat.router.mycat1xfunction.PartitionByModname: PartitionByModproperties:count: '4'columnName: 'id'
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByMod","name": "PartitionByMod","properties": {"count": "4","columnName": "id"}}
"xxx":{"createTableSQL":"......","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByMod","properties":{"count":"80","columnName":"id"},"ranges":{}},"partition":{"schemaNames":"aaaa","tableNames":"xxxx_$0-79","targetNames":"prototype"}},
count是partition(dataNode)的数量
此种配置非常明确即根据 id 进行十进制求模预算，相比固定分片 hash，此种在批量插入时可能存在批量插入单
事务插入多数据分片，增大事务一致性难度。
按日期（天）分片
第一种方式,对应1.6的自然日分片
v1.21- release-3-22之后支持
/*+ mycat:createTable{"schemaName":"db1","shardingTable":{"createTableSQL":"CREATE TABLE db1.`sharding` (\n `id` bigint NOT NULL AUTO_INCREMENT,\n `user_id` varchar(100) DEFAULT NULL,\n `create_time` date DEFAULT NULL,\n `fee` decimal(10,0) DEFAULT NULL,\n `days` int DEFAULT NULL,\n `blob` longblob,\n PRIMARY KEY (`id`),\n KEY `id` (`id`)\n) ENGINE=InnoDB DEFAULT CHARSET=utf8","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByDateEx","properties":{"beginDate":"2021-01-01","dateFormat":"yyyy-MM-dd","partionDay":10,"columnName":"create_time","naturalDay":"0"}},"partition":{"schemaNames":"db1","tableNames":"sharding_$0-72","targetNames":"prototype"}},"tableName":"sharding"} */;//explainINSERT INTO db1.sharding VALUES (null, 'test1', '2021-01-01', 2, 3, NULL);//路由到db1.sharding_0explainINSERT INTO db1.sharding VALUES (null, 'test1', '2022-01-01', 2, 3, NULL);//路由到db1.sharding_36
第二种方式,只支持一年之内的日期分片
- clazz: io.mycat.router.mycat1xfunction.PartitionByDatename: PartitionByDateproperties:dateFormat: yyyy-MM-ddbeginDate: 2014-01-01endDate: 2014-01-02partionDay: 10
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByDate","name": "PartitionByDate","properties": {"dateFormat": "yyyy-MM-dd","beginDate": "2014-01-01T00:00:00.000Z","endDate": "2014-01-02T00:00:00.000Z","partionDay": 10,"columnName": "id"}
dateFormat ：日期格式
beginDate ：开始日期
endDate：结束日期
partionDay ：分区天数，即默认从开始日期算起，分隔 10 天一个分区
分片数量=(endDate - beginDate)/partionDay
如果配置了 endDate 则代表数据达到了这个日期的分片后后循环从开始分片插入
https://github.com/MyCATApache/Mycat2/blob/62b9899938637b0c75176dfb71829eebbd8ba647/router/src/test/java/io/mycat/router/mycat1xfunction/PartitionByDateTest.java
取模范围约束
- clazz: io.mycat.router.mycat1xfunction.PartitionByPatternname: PartitionByPatternproperties:patternValue: 256defaultNode: -1columnName: 'id'ranges:1-32: 033-64: 1
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByPattern","name": "PartitionByPattern","properties": {"patternValue": 256,"defaultNode": -1,"columnName": "id"},"ranges": {"1-32": 0,"33-64": 1}}
patternValue: 即求模基数，
defaoultNode默认节点，如果分片值转换数字失败,则不会按照求模运算,计算结果是默认节点
1-32 即代表 id%256 后分布的范围，如果在 1-32 则在分区 1，其他类推，如果 id 非数据，则
会分配在 defaoultNode 默认节点
/*+ mycat:createTable{"schemaName":"db1","shardingTable":{"createTableSQL":"create table address4(\nid int(11) NOT NULL AUTO_INCREMENT,\nuser_id int(11) ,\nuser_name varchar(128), \nPRIMARY KEY (`id`) \n)ENGINE=InnoDB DEFAULT CHARSET=utf8 ","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByPattern","properties":{"type":"Integer","patternValue": 256,"defaultNode": 1,"columnName": "id"},"ranges": {"1-32": 0,"33-64": 1,"65-72": 2}},"partition":{"schemaNames":"db3","tableNames":"addr4_$0-2","targetNames":"prototype",}},"tableName":"address4"} */;
2022-1-6后支持在properties写上mapFile属性指定ranges范围
截取数字做 hash 求模范围约束
- clazz: io.mycat.router.mycat1xfunction.PartitionByPrefixPatternname: PartitionByPrefixPatternproperties:patternValue: 256prefixLength: 5columnName: 'id'ranges:1-4: 05-8: 19-12: 213-16: 317-20: 421-24: 525-28: 629-32: 70-0: 7
{"clazz": "io.mycat.router.mycat1xfunction.PartitionByPrefixPattern","name": "PartitionByPrefixPattern","properties": {"patternValue": 256,"prefixLength": 5,"columnName": "id"},"ranges": {"1-4": 0,"5-8": 1,"9-12": 2,"13-16": 3,"17-20": 4,"21-24": 5,"25-28": 6,"29-32": 7,"0-0": 7}}
"shardingTables":{"topic":{"createTableSQL":"CREATE TABLE ....","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByPrefixPattern","properties":{"prefixLength":13,"patternValue":32,"storeNum":1,"columnName":"id"},"ranges":{"0-31" : 0}},"partition":{"schemaNames":"....","tableNames":"..._$0-31","targetNames":"prototype"}}}
patternValue:即求模基数
prefixLength:ASCII 截取的位数
1-32 即代表 分片值%256 后分布的范围，如果在 1-32 则在分区 1，其他类推
采取的是将列种获取前 prefixLength 位列所有 ASCII 码的和进行求模
sum%patternValue ,获取的值，在范围内的分片数，
/*+ mycat:createTable{"schemaName":"db1","shardingTable":{"createTableSQL":"create table sharding(\nid int(11) NOT NULL AUTO_INCREMENT,\nuser_id int(11) ,\nuser_name varchar(128), \nPRIMARY KEY (`id`), \n GLOBAL INDEX `g_i_user_id`(`user_id`) COVERING (`user_name`) dbpartition by btree(`user_id`) \n)ENGINE=InnoDB DEFAULT CHARSET=utf8 ","function":{"clazz":"io.mycat.router.mycat1xfunction.PartitionByPrefixPattern","properties":{"prefixLength":3,"patternValue":72,"columnName":"id"},"ranges":{"0-31" : 0,"32-64" : 1,"65-72" : 2,}},"partition":{"schemaNames":"testSchema","tableNames":"sharding_$0-3","targetNames":"prototype"}},"tableName":"sharding"} */;
2022-1-6后支持在properties写上mapFile属性指定ranges范围
应用指定
- clazz: io.mycat.router.mycat1xfunction.PartitionDirectBySubStringname: PartitionDirectBySubStringproperties:startIndex: 0size: 2partitionCount: 8defaultNode: 0columnName: 'id'
{"clazz": "io.mycat.router.mycat1xfunction.PartitionDirectBySubString","name": "PartitionDirectBySubString","properties": {"startIndex": 0,"size": 2,"partitionCount": 8,"defaultNode": 0,"columnName": "id"}
此规则是在运行阶段有应用自主决定路由到那个分片。
此方法为直接根据字符子串（必须是数字）计算分区号（由应用传递参数，显式指定分区号）。
例如 id=05-100000002
在此配置中代表根据 id 中从 startIndex=0，开始，截取 size=2 位数字即 05，05 就是获取的分片，
默认分配到 defaultNode
截取数字 hash 解析(sharding-string-hash)
- clazz: io.mycat.router.mycat1xfunction.PartitionByString
name: PartitionByString
properties:
partitionLength: 128,128,256
partitionCount: 2,2,2
hashSlice: -1:0
columnName: 'id'
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByString",
"name": "PartitionByString",
"properties": {
"partitionLength": "128,128,256",
"partitionCount": "2,2,2",
"hashSlice": -60,
"columnName": "id"
}
}
hashSlice : 配置 分片键值 范围截取规则； （hash 预算位 格式为 start:end）
partitionLength 每个分区占用长度
partitionCount 分区数量
hashSlice 示例： 假设 user_id = 11356789,以下规则截取出的结果如下所示
配置中的特殊值（0 means str.length(), -1 means str.length()-1）
“2” -> (0,2) -> 1
“1:2” -> (1,2) -> 11
“1:” -> (1,0) -> 1356789
“-1:” -> (-1,0) -> 9
“:-1” -> (0,-1) -> 1135678
“:” -> (0,0) -> 11356789
分区范围： 需要 partitionLength 和 partitionCount 配合使用
partitionLength : 128,128,256
partitionCount : 2,2,2
所有分区范围必须是 0~1024： 如上所示 128_2 + 128_2 +256 _2 = 1024 (sum(partitionLength[i]_partitionCount[i]))
如上配置表示： 总共有 6 个分区,每个分区范围如下所示
1， 2， 3， 4， 5， 6
0~128,128~256,256-384,384-512,512-768,768-1024
一致性 hash
一致性 hash 预算有效解决了分布式数据的扩容问题
- clazz: io.mycat.router.mycat1xfunction.PartitionByMurmurHash
name: PartitionByMurmurHash
properties:
seed: 0
count: 2
virtualBucketTimes: 160
columnName: 'id'
ranges:
1: 0 # 键:权重
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByMurmurHash",
"name": "PartitionByMurmurHash",
"properties": {
"seed": 0,
"count": 2,
"virtualBucketTimes": 160,
"columnName": "id"
},
"ranges": {
"1": 0
}
}
seed :默认是 0
count:要分片的数据库节点数量， 必须指定， 否则没法分片
virtualBucketTimes :一个实际的数据库节点被映射为这么多虚拟
节点， 默认是 160 倍， 也就是虚拟节点数是物理节点数的 160 倍
ranges:
节点的权重， 没有指定权重的节点默认是 1。 以 properties 文件的格式填写， 以从 0 开始到 count-1 的整数值也就是节点索引为 key， 以节点权重值为值。 所有权重值必须是正整数， 否则以 1 代替
按单月小时拆分
此规则是单月内按照小时拆分，最小粒度是小时，可以一天最多 24 个分片，最少 1 个分片，一个月完后下月
从头开始循环。
每个月月尾，需要手工清理数据。
- clazz: io.mycat.router.mycat1xfunction.PartitionByLatestMonth
name: PartitionByLatestMonth
properties:
dateFormat: yyyymmddHH
splitOneDay: 24
columnName: 'id'
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByMurmurHash",
"name": "PartitionByMurmurHash",
"properties": {
"seed": 0,
"count": 2,
"virtualBucketTimes": 160,
"columnName": "id"
},
"ranges": {
"1": 0
}
}
范围求模分片
先进行范围分片计算出分片组，组内再求模
优点可以避免扩容时的数据迁移，又可以一定程度上避免范围分片的热点问题
综合了范围分片和求模分片的优点，分片组内使用求模可以保证组内数据比较均匀，分片组之间是范围分片可以兼顾范围查询。
最好事先规划好分片的数量，数据扩容时按分片组扩容，则原有分片组的数据不需要迁移。由于分片组内数据比较均匀，所以分片组内可以避免热点数据问题。
- clazz: io.mycat.router.mycat1xfunction.PartitionByRangeMod
name: PartitionByRangeMod
properties:
defaultNode: -1
columnName: 'id'
ranges:
0-200M: 5
200M-400M: 1
400M-600M: 4
600M-800M: 4
800M1-1000M: 6
[
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByRangeMod",
"name": "PartitionByRangeMod",
"properties": {
"defaultNode": -1,
"columnName": "id"
},
"ranges": {
"0-200M": 5,
"200M-400M": 1,
"400M-600M": 4,
"600M-800M": 4,
"800M1-1000M": 6
}
}
]
/*+ mycat:createTable{
"schemaName":"testSchema",
"shardingTable":{
"createTableSQL":"create table sharding(\nid int(11) NOT NULL AUTO_INCREMENT,\nuser_id int(11) ,\nuser_name varchar(128), \nPRIMARY KEY (`id`), \n GLOBAL INDEX `g_i_user_id`(`user_id`) COVERING (`user_name`) dbpartition by btree(`user_id`) \n)ENGINE=InnoDB DEFAULT CHARSET=utf8 ",
"function":{
"clazz":"io.mycat.router.mycat1xfunction.PartitionByRangeMod",
"properties":{
"defaultNode":-1,
"columnName":"id"
},
"ranges":{
"0-1":1,
"1-300":2
}
},
"partition":{
"schemaNames":"testSchema",
"tableNames":"sharding",
"targetNames":"dw0,dw1"
}
},
"tableName":"sharding"
} */;
日期范围 hash 分片
思想与范围求模一致，当由于日期在取模会有数据集中问题，所以改成 hash 方法。
先根据日期分组，再根据时间 hash 使得短期内数据分布的更均匀
优点可以避免扩容时的数据迁移，又可以一定程度上避免范围分片的热点问题
要求日期格式尽量精确些，不然达不到局部均匀的目的
- clazz: io.mycat.router.mycat1xfunction.PartitionByRangeDateHash
name: PartitionByRangeDateHash
properties:
dateFormat: yyyy-MM-dd HH:mm:ss
partionDay: 3
groupPartionSize: 6
columnName: 'id'
beginDate: '2018-01-31 00:00:00'
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByRangeDateHash",
"name": "PartitionByRangeDateHash",
"properties": {
"dateFormat": "yyyy-MM-dd HH:mm:ss",
"partionDay": 3,
"groupPartionSize": 6,
"columnName": "id",
"beginDate": '2018-01-31 00:00:00'
}
}
partionDay 代表多少天分一个分片
groupPartionSize 代表分片组的大小
冷热数据分片
根据日期查询日志数据 冷热数据分布 ，最近 n 个月的到实时交易库查询，超过 n 个月的按照 m 天分片。
- clazz: io.mycat.router.mycat1xfunction.PartitionByHotDate
name: PartitionByHotDate
properties:
dateFormat: yyyy-MM-dd
lastDay: 10
partionDay: 30
columnName: 'id'
{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByHotDate",
"name": "PartitionByHotDate",
"properties": {
"dateFormat": "yyyy-MM-dd",
"lastDay": 10,
"partionDay": 30,
"columnName": "id"
}
}
/*+ mycat:createTable{
"schemaName":"db1",
"shardingTable":{
"createTableSQL":"CREATE TABLE db1.`sharding` (\n `id` bigint NOT NULL AUTO_INCREMENT,\n `user_id` varchar(100) DEFAULT NULL,\n `create_time` date DEFAULT NULL,\n `fee` decimal(10,0) DEFAULT NULL,\n `days` int DEFAULT NULL,\n `blob` longblob,\n PRIMARY KEY (`id`),\n KEY `id` (`id`)\n) ENGINE=InnoDB DEFAULT CHARSET=utf8",
"function":{
"clazz":"io.mycat.router.mycat1xfunction.PartitionByHotDate",
"properties":{
"dateFormat":"yyyy-MM-dd",
"lastTime":90,
"partionDay":180,
"columnName":"create_time"
}
},
"partition":{
"schemaNames":"db1",
"tableNames":"sharding_0,sharding_1",
"targetNames":"c0"
}
},
"tableName":"sharding"
} */;
要求分片字段是时间类型
它是系统时间与分片值的差值是否在90天内,是90天内,去到0节点,否则是旧历史分区 ,因为系统时间每天都在变化,计算算法与系统时间有关,所以昨天能查询的值(临界值,快过期的值,路由去0节点),今天可能查询不到(路由不去0节点,路由到历史分区节点),如果没有没有迁移程序配合会出现不一致的情况,但是对于热点数据,总是查询90天之内的数据是正确的.
需要有一个工具,把前90天那一天的数据挪到分表
基本形式是
分一个新表,一个历史表,新表数据随着时间迁移到历史表
自然月分片
- clazz: io.mycat.router.mycat1xfunction.PartitionByMonth
name: PartitionByMonth
properties:
formatter: yyyy-MM-dd
beginDate: 2014-01-01
endDate: 2014-12-01
columnName: 'id'
"function":{
"clazz": "io.mycat.router.mycat1xfunction.PartitionByMonth",
"name": "PartitionByMonth",
"properties": {
"dateFormat": "yyyy-MM-dd hh:mm:ss",
"beginDate": "2021-01-01 00:00:00",
"endDate": "",
"columnName": "create_time"
},
"ranges":{}
},
/*+ mycat:createTable{
"schemaName":"xxxx",
"shardingTable":{
"createTableSQL":"xxxx",
"function":{
"clazz":"io.mycat.router.mycat1xfunction.PartitionByMonth",
"properties":{
"beginDate":"2019-01-01",
"endDate":"2099-12-01",
"dateFormat":"yyyy-MM-dd",
"columnName":"date"
}
},
"partition":{
"schemaNames":"c",
"tableNames":"xxx_$0-11",
"targetNames":"prototype"
}
},
"tableName":"xxx"
} */;
columns： 分片字段，字符串类型
dateFormat ： 日期字符串格式,默认为 yyyy-MM-dd
beginDate ： 开始日期，无默认值
endDate：结束日期，无默认值
节点从 0 开始分片
使用场景：
场景 1：
默认设置； 节点数量必须是 12 个， 从 1 月~12 月
"2014-01-01" = 节点 0
"2013-01-01" = 节点 0
"2018-05-01" = 节点 4
"2019-12-01" = 节点 11
场景 2：
beginDate = “2017-01-01”
该配置表示”2017-01 月”是第 0 个节点， 从该时间按月递增， 无最大节点
"2014-01-01" = 未找到节点
"2017-01-01" = 节点 0
"2017-12-01" = 节点 11
"2018-01-01" = 节点 12
"2018-12-01" = 节点 23
场景 3：
beginDate = “2015-01-01”sEndDate = “2015-12-01”
该配置可看成与场景 1 一致； 场景 1 的配置效率更高
"2014-01-01" = 节点 0
"2014-02-01" = 节点 1
"2015-02-01" = 节点 1
"2017-01-01" = 节点 0
"2017-12-01" = 节点 11
"2018-12-01" = 节点 11
该配置可看成是与场景 1 一致
场景 4：
beginDate = “2015-01-01”sEndDate = “2015-03-01”
该配置标识只有 3 个节点； 很难与月份对应上； 平均分散到 3 个节点上
若有收获，就点个赞吧
0 人点赞
