lenmore

没必要 redis 。几千万数据，建好索引，MySQL 没有任何问题。

推荐 3 个节点
但是预算实在紧张，可以考虑 1 主 1 从 + 1 个 Arbiter ，参考： https://www.mongodb.com/docs/manual/core/replica-set-arbiter/

MySQL 可以试下这个，以前写的应付检查的……
导出成 excel ，去掉表头就行了。

SET @table_schema='test';

SELECT column_name, DATA_TYPE, COLUMN_COMMENT
FROM (
SELECT table_name, '' AS column_name, '' AS DATA_TYPE, '' AS COLUMN_COMMENT, -4 AS ORDINAL_POSITION FROM information_schema.tables WHERE table_schema=@table_schema
UNION
SELECT table_name, CONCAT('表名：', table_name) AS column_name, '' AS DATA_TYPE, '' AS COLUMN_COMMENT, -3 AS ORDINAL_POSITION FROM information_schema.tables WHERE table_schema=@table_schema
UNION
SELECT table_name, CONCAT('用途：', table_comment) AS column_name, '' AS DATA_TYPE, '' AS COLUMN_COMMENT, -2 AS ORDINAL_POSITION FROM information_schema.tables WHERE table_schema=@table_schema
UNION
SELECT table_name, '字段名' AS column_name, '字段类型' AS DATA_TYPE, '描述' AS COLUMN_COMMENT, -1 AS ORDINAL_POSITION FROM information_schema.tables WHERE table_schema=@table_schema
UNION
SELECT table_name, column_name, CONCAT(DATA_TYPE, CASE WHEN CHARACTER_MAXIMUM_LENGTH IS NOT NULL THEN CONCAT('(', CHARACTER_MAXIMUM_LENGTH, ')') WHEN NUMERIC_PRECISION IS NOT NULL THEN CONCAT('(',NUMERIC_PRECISION, CASE WHEN NUMERIC_SCALE>0 THEN CONCAT(',',NUMERIC_SCALE) ELSE '' END , ')') WHEN DATETIME_PRECISION > 0 THEN CONCAT('(', DATETIME_PRECISION, ')') ELSE '' END) AS DATA_TYPE, COLUMN_COMMENT,ORDINAL_POSITION FROM information_schema.columns WHERE table_schema=@table_schema
) AS t
ORDER BY table_name, ORDINAL_POSITION;

用 percent_rank 窗口函数，然后对 pct 分组，统计出各组的平均分。

e.g.
select floor(pct*100 /25) as grp, avg(n)
from (
select n, percent_rank() over(order by n) as pct
from public.global_number order by n
) as t
group by floor(pct*100 /25)
order by grp

谢谢楼主的提醒，已全部迁移到 KeePassXC+浏览器插件

也用 HackBrowserData 工具测试了，的确可以导出 Chrome/Edge/Firefox 的密码
Firefox 加了主密码后这个工具导不出了，
Edge 的主密码是摆设

1 、xtrabackup 每日全量备份
2 、mysqlbinlog 可以做准实时 binlog backup 。https://dev.mysql.com/doc/refman/5.6/en/mysqlbinlog-backup.html

注意文件都不要放本地

价格表做点字段冗余，就不需要 join 这一堆表了。