skeema / skeema / 11596597080

92%

push

github

docker: disable redo log on ephemeral MySQL 8.0.21+ instances

This commit improves performance of ephemeral Dockerized MySQL 8.0.21+
servers used by Skeema's workspace=docker option, as well as by Skeema's
integration test suites. All such Dockerized database servers will now have
redo logging disabled automatically. Unlike previous Docker-related config
change commits, this is applied even to pre-existing Skeema containers, since
it uses ALTER INSTANCE rather than a command-line server variable.

InnoDB's redo log is its primary crash recovery mechanism, which is normally
essential for production databases, but isn't necessary for ephemeral DBs
which can be easily destroyed and recreated at will. Disabling the redo log
improves MySQL 8.x/9.x integration test execution time by over 20%.

This commit was inspired by an excellent blog post by Jean-François Gagné:
https://jfg-mysql.blogspot.com/2024/10/impressed-by-disabling-innodb-redo-logs.html

When using workspace=docker with the default setting of docker-cleanup=none
(which leaves the containers running after Skeema exits), the redo log is re-
enabled just before Skeema exits. This prevents the containers from breaking
if a host crash occurs in the future.

When using workspace=docker with docker-cleanup=destroy on a native Linux
host, the performance gains from this commit may be negligible, since Skeema
automatically uses a tmpfs mount for the data directory in that situation.
See commit 766efc3.

24 of 26 new or added lines in 2 files covered. (92.31%)

10059 of 10787 relevant lines covered (93.25%)

1.11 hits per line