ShardingProxy(ShardingJDBC)自定义分表规则

一、场景

在使用ShardingJDBC或ShardingProxy时,有些表按照取模/hash无法满足分库分表规则,这个时候可以自己实现分库分表的策略.

下面按照时间分表,orgId分库.

二、代码编写

1.创建maven工程，引入pom依赖。

<dependencies>
    <dependency>
        <groupId>org.apache.shardingspheregroupId>
        <artifactId>sharding-core-apiartifactId>
        <version>4.1.1version> //自己引入的对应版本
    dependency>
dependencies>

2.范围分片算法.

• 实现RangeShardingAlgorithm接口
• 重写doSharding方法

/**
 * @BelongsProject: DemoCode
 * @BelongsPackage: 
 * @Author: hef
 * @CreateTime: 2021-07-09 17:22
 * @Description: 
 */
public class CustomTableRangeShardingAlgorithm implements RangeShardingAlgorithm<String> {
    public CustomTableRangeShardingAlgorithm() {
    }


    @Override
    public Collection<String> doSharding(Collection<String> collection, RangeShardingValue<String> rangeShardingValue) {
        Range<String> ranges = rangeShardingValue.getValueRange();
        LocalDateTime start = null;
        Integer startYM = null;
        if (ranges.hasLowerBound()) {
            String lower = ranges.lowerEndpoint();
            start = LocalDatetimeUtil.parse(lower);
            startYM = Integer.valueOf(start.format(DateTimeFormatter.ofPattern("yyyyMM")));
        }
        LocalDateTime end = null;
        Integer endYM = null;
        if (ranges.hasUpperBound()) {
            String upper = ranges.upperEndpoint();
            end = LocalDatetimeUtil.parse(upper);
            endYM = Integer.valueOf(end.format(DateTimeFormatter.ofPattern("yyyyMM")));
        }
        Collection<String> tables = new LinkedHashSet<>();
        if (judgmentTime(start, end)) {
            for (String c : collection) {
                String substring = c.substring(c.length() - ShardingConstant.DATE_SHARDING_TABLE_FORMAT.length());
                Integer cMonth = Integer.valueOf(LocalDatetimeUtil.parse(substring).format(DateTimeFormatter.ofPattern("yyyyMM")));
                if (start != null && end != null) {
                    if (cMonth >= startYM && cMonth <= endYM) {
                        tables.add(c);
                    }
                } else if ((start != null && cMonth >= startYM) || (end != null && cMonth <= endYM)) {
                    tables.add(c);
                }
            }
        }
        return tables;
    }

    public boolean judgmentTime(LocalDateTime start, LocalDateTime end) {
        if (start == null && end == null) {
            throw new IllegalArgumentException("no valid time");
        }
        if (start != null && end != null) {
            return start.getNano() <= end.getNano();
        }
        return true;
    }
}

3、精确分片算法。

• 实现PreciseShardingAlgorithm接口
• 重写doSharding方法

/**
 * @BelongsProject: DemoCode
 * @BelongsPackage: 
 * @Author: hef
 * @CreateTime: 2021-07-09 17:27
 * @Description: 
 */
public class CustomTableShardingAlgorithm implements PreciseShardingAlgorithm<String> {

    public CustomTableShardingAlgorithm() {
    }

    @Override
    public String doSharding(Collection<String> collection, PreciseShardingValue<String> preciseShardingValue) {
        String value = preciseShardingValue.getValue();
        LocalDateTime createTime = LocalDatetimeUtil.parse(value);
        String timeValue = createTime.format(DateTimeFormatter.ofPattern(ShardingConstant.DATE_SHARDING_TABLE_FORMAT));

        String columnName = preciseShardingValue.getColumnName();
        // 需要分库的逻辑表
        String table = preciseShardingValue.getLogicTableName();
        if (timeValue == null || timeValue.length() == 0) {
            throw new UnsupportedOperationException(columnName + ":列，分表精确分片值为NULL;");
        }
        for (String each : collection) {
            if (each.startsWith(table)) {
                return table + "_" + timeValue;
            }
        }
        return table;
    }
}

4、工具类

/**
 * @BelongsProject: DemoCode
 * @BelongsPackage: 
 * @Author: hef
 * @CreateTime: 2021-07-12 10:50
 * @Description: 
 */
public class LocalDatetimeUtil {

    private static List<DateTimeFormatter> formatList = null;

    static {
        formatList = new ArrayList<>();
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.S"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.SS"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss.SSS"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss.S"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss.SS"));
        formatList.add(DateTimeFormatter.ofPattern("yyyy-MM-dd'T'HH:mm:ss.SSS"));
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern("yyyy-MM-dd")
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern("yyyy-MM")
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern("MM-dd")
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
        formatList.add(DateTimeFormatter.ofPattern("yyyy/MM/dd"));
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern("yyyy/MM/dd")
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
        formatList.add(DateTimeFormatter.ofPattern("yyyy/MM/dd HH:mm:ss"));
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern("yyyyMMdd")
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
        formatList.add(DateTimeFormatter.ofPattern("yyyyMMdd HH:mm:ss"));
        formatList.add(DateTimeFormatter.ofPattern("yyyyMMddHHmmss"));
        formatList.add(DateTimeFormatter.ofPattern("yyyyMMddHHmmssSSS"));
        formatList.add(
                new DateTimeFormatterBuilder()
                        .appendPattern(ShardingConstant.DATE_SHARDING_TABLE_FORMAT)
                        .parseDefaulting(ChronoField.MONTH_OF_YEAR, 1)
                        .parseDefaulting(ChronoField.DAY_OF_MONTH, 1)
                        .parseDefaulting(ChronoField.HOUR_OF_DAY, 0)
                        .toFormatter());
    }

    /**
     * 格式化时间
     *
     * @param dateStr
     * @return
     */
    public static LocalDateTime parse(String dateStr) {
        LocalDateTime time = null;
        for (DateTimeFormatter formatter : formatList) {
            try {
                time = LocalDateTime.parse(dateStr, formatter);
                break;
            } catch (Exception e) {
                continue;
            }
        }
        if (time == null) {
            throw new IllegalArgumentException("wrong time format!:" + dateStr);
        }
        return time;
    }
}

/**
 * @BelongsProject: DemoCode
 * @BelongsPackage: 
 * @Author: hef
 * @CreateTime: 2021-07-12 11:40
 * @Description: 
 */
public class ShardingConstant {

    /**
     * 日期分表格式
     */
    public static final String DATE_SHARDING_TABLE_FORMAT = "yyyy_MM";
}

三、ShardingProxy配置。

注意要把刚刚写好的maven工程打成jar包导入到proxy的lib目录下面.

shardingRule:
  tables:
    #记录表
    capture_info:
      actualDataNodes: face_${1..16}.capture_info_${2021..2022}_0$->{1..9},face_${1..16}.capture_info_${2021..2022}_$->{10..12}
      tableStrategy:
        #注意这一段配置
        standard:
          #分表键
          shardingColumn: capture_time
          #范围分表规则--->指定刚刚编写的类路径
          rangeAlgorithmClassName: com.rock.sharding.config.CustomTableRangeShardingAlgorithm
          #精确分表规则--->指定刚刚编写的类路径
          preciseAlgorithmClassName: com.rock.sharding.config.CustomTableShardingAlgorithm
      databaseStrategy:
        inline:
          shardingColumn: org_id
          algorithmExpression: face_${org_id%16+1}

配置完成后启动proxy,然后连接到proxy后执行建表语句,则会自动创建对应格式的表;

四、额外的话

• 使用ShardingProxy启动的代理数据库无法通过Navicat和SQLYog无法连接的时候可以试试idea自带的数据库连接工具或者使用DataGrip和命令行模式试试。
• 如果担心proxy服务不稳定可以使用Nginx做个负载均衡。
• 此文章仅代表博主在使用ShardingProxy过程中遇到的需求然后记录下，不保证观点完全准确。

参考文章：https://www.cnblogs.com/liran123/p/14191240.html