The MySQL documentation is not always right

Let me premise this post with the statement I think the MySQL documentation is an excellent and highly accurate resource. I think the MySQL docs team do a great job, however like software and people, documentation is not perfect.

As members of the MySQL community you can always contribute to improve the process by reading the documentation and logging any issues as Documentation Bugs.

Some time ago in a discussion with a friend and colleague, we were talking about changes in historical defaults that had been improved finally in MySQL 5.4 The specific discussion was on the new default innodb_buffer_pool_size and we both agreed it increased significantly. One said 1GB, the other said 128MB. Who was right? Well we both were, and we were both inaccurate depending on versions.

Referencing the 5.4 Manual in InnoDB Startup Options and System Variables the current value for Linux is 128M, but for Windows it’s 1GB.

However I was confident I was told in a presentation, perhaps even the keynote the value was 1GB. Firing up my server and seeing the original version I used of 5.4.0 (which was not available on Windows) we find that the default for Linux was 1GB at some time, i.e. the first release.

mysql> show global variables like 'innodb_buffer_pool_size';
| Variable_name           | Value      |
| innodb_buffer_pool_size | 1073741824 |
1 row in set (0.00 sec)

mysql> select 1073741824/1024/1024 as MB;
| MB            |
| 1024.00000000 |
1 row in set (0.00 sec)

mysql> show global variables like 'version%';
| Variable_name           | Value                        |
| version                 | 5.4.0-beta                   |
| version_comment         | MySQL Community Server (GPL) |
| version_compile_machine | x86_64                       |
| version_compile_os      | unknown-linux-gnu            |
4 rows in set (0.00 sec)

I’m not trying to nit pick here, I’m highlighting that MySQL is a evolving product with many different versions and architectures. It’s our job of the MySQL community to help make the documentation the best for all readers. In this above case I’ve not logged the issue, because 5.4 is a defunct product, however if you want an example of how I identified a problem, provided a test case, and saw that my contribution was reviewed, verified and implemented check out Bug #51739 –core-file is not default TRUE (incorrect docs).

In conclusion, always read the documentation but pay special attention to the current version that matches the documentation, and the version you are actually running. Defaults change between versions, e.g. innodb_thread_concurrency is a complex example, and I’ve been caught with a large enterprise client with assuming the default of a Connector/J options as true, when it was in 5.0.6, but in 5.0.5 the version the client was running it was false.

An old saying, “trust by verify” is a good motto to consider.


  1. says

    On a related note to this argument is the notion of accepting defaults on an important daemon. It’s not always reasonable due to the large amount of configuration options some programs provide, but I’ve taken to always explicitly defining defaults as I come across them.

    Especially with things like Mysql/Apache/script interpreter of choice, taking a day to go through and explicitly define all the defaults will save many headaches relating to upgrading and ensuring the defaults are actually what the stale documentation claims they are. I’ll usually make a comment noting that this was a “default value”, so other viewers don’t assume something was set for a local reason (and may be a bit more leery of changing it).