共计 3018 个字符,预计需要花费 8 分钟才能阅读完成。
导读 | 今天的分享是再批 json, 去年分享过因为 mysql json 导致的故障,今天的 case 其实是去年的姊妹篇,原理一模一样。上一篇弱智的 MySQL NULL, 居然有小伙伴留言说,在业务中依赖 NULL 使联合索引不唯一的特性,比如有的用户就要多条记录,有的仅一条。我看了差点一口老血喷出来,把业务逻辑耦合在 DB 中这样真的合适嘛? 要是外包另当别论,正常项目谁接手谁倒霉。 |
今天的分享是再批 json, 去年分享过因为 mysql json 导致的故障,今天的 case 其实是去年的姊妹篇,原理一模一样。有两个原因不建议用 json:
上面提到的两点有争议? 有争议就对了,一致认同是垃圾的东西谁会讨论它呢?
JSON 有两种表示方法:文本可读的在 mysql 中对应 json_dom.cc, binary 二进制表示的对应 json_binary.cc
If the value is a JSON object, its binary representation will have a
header that contains:
- the member count
- the size of the binary value in bytes
- a list of pointers to each key
- a list of pointers to each value
The actual keys and values will come after the header, in the same
order as in the header.
Similarly, if the value is a JSON array, the binary representation
will have a header with
- the element count
- the size of the binary value in bytes
- a list of pointers to each value
源码中注释也写的比较清楚,二进制分成两部分 header + element. 实际上 mysql 只是 server 识别了 json, 各个存储引擎仍存储的二进制 blob
换句话说,底层引擎对 json 是无感知的,就是一条数据而己
json-function-reference[1] 官方有好多在 server 层操作 json 的方法,感兴趣的可以看一下
MySQL Client 读取 json 时是 json_dom 调用 wrapper_to_string 方法,序列化成可读格式数据
写入 json 时,是由 json_binary 调用 serialize_json_value 方法,序列化成上面图表示的 binary 数据,然后由引擎层存储成 blob 格式
去年故障也有服务端的问题:加载单条数据失败主动 panic, 坑人不浅 (理由是数据不一致,宁可不对外提供服务,问题是那条数据恰好是重不重要的一类)。所以这个故事告诉我们: 在线服务的可用性,远高于数据一致性
慢的原因是 wrapper_to_string 遇到 json array 特别多的情况下反复 mem_realloc 创建内存空间,导致性能下降
其实去年没有 fix 完整,最近发现写入也有类似问题,只不过是 serialize_json_value 写入存储引擎前反复 mem_realloc 造成超时。这时前端页面发现写入超时了,(人工) 重试继续写入 json 数据
恰好赶上联合索引中有 NULL 字段,由此引出了唯一索引不唯一的现象。那怎么解决呢? 前端按钮 cooldown 治标不治本,sql 执行 12s 前端肯定又点击提交了,治本还得升级 mysql 8.0 并且移除 NULL 字段, 那会不会又引入其它问题呢?
项目初期做了错误的决定,后人很容易买单。希望我们踩到的坑,能让你决定使用 json 前犹豫几秒钟 ^^
在测试机上发现 8.0 是 ok 的,没有性能问题,查看提交的 commit, 2016 年就有人发现并 fix 了,不知道有没有 back port 到 mysql 5.7 那几个版本
commit a2f9ea422e4bdfd65da6dd0c497dc233629ec52e
Author: Knut Anders Hatlen
Date: Fri Apr 1 12:56:23 2016 +0200
Bug#23031146: INSERTING 64K SIZE RECORDS TAKE TOO MUCH TIME
If a JSON value consists of a large sub-document which is wrapped in
many levels of JSON arrays or objects, serialization of the JSON value
may take a very long time to complete.
This is caused by how the serialization switches between the small
storage format (used by documents that need less than 64KB) and the
large storage format. When it detects that the large storage format
has to be used, it redoes the serialization of the current
sub-document using the large format. But this re-serialization has to
be redone again when the parent of the sub-document is switched from
small format to large format. For deeply nested documents, the inner
parts end up getting re-serializing again and again.
This patch changes how the switch between the formats is done. Instead
of starting with re-serializing the inner parts, it now starts with
the outer parts. If a sub-document exceeds the maximum size for the
small format, we know that the parent document will exceed it and need
to be re-serialized too. Re-serializing an inner document is therefore
a waste of time if we haven't already expanded its parent. By starting
with expanding the outer parts of the JSON document, we avoid the
wasted work and speed up the serialization.
参考资料
[1]json-function-reference: https://dev.mysql.com/doc/refman/5.7/en/json-function-reference.html