Description:
Documentation exists on https://dev.mysql.com/doc/mysql-sourcebuild-excerpt/8.0/en/installing-source-distribution.... (and for newer versions) on how to build MySQL. However, the actual hardware resources required to build the MySQL server these days is not very clearly explained even approximately.
Clearly there are multiple factors involved, there is lots of different hardware and things will change over time yet it's convenient to have an idea of the approximate build resources required to build MySQL.
That may be more important if building MySQL using containers where often container resources may not be unlimited.
Adhoc observations for me indicates that as of August 2025 it appears that more than 8GB of RAM is needed for me to be able to build at least the MySQL rpms on RHEL equivalent OS versions. Having 8 GB seems to be insufficient.
Clearly this also depends on the specific compiler being used, though for rpm builds the community rpms are built with a specific gcc toolset compiler.
Also this may be influenced significantly by the number of processors seen by make / cmake and its ability to do things in parallel, so most likely seeing more processors will trigger more memory to be used.
In my case I've tried building MySQL (rpm versions specifically) on small NUC type hardware with ~8 cores and this seems to be successful on machines with ~16 GB or larger.
The reason for caring about this is that if you attempt to build MySQL and have insufficient memory the build process will appear to start properly and then may fail at some undefined place later. If building using a container the container may die immediately after running out of memory so determining the cause may be less obvious as the errors reported during the build may not be clear and it may be necessary to inspect the container infrastructure to see this.
How to repeat:
See above.
Suggested fix:
Provide some sort of rough estimation of how many resources are needed to build MySQL by major version, so 8.4 LTS / 9.X Innovation or LTS on a given hardware setup. That hardware may become outdated over time but at least it gives some sort of benchmark to compare with.
Do this if possible for popular architectures, unless the variation seen is minimal and indicate if providing more "cores" to the build process affects the number determined above or not.
