We run all queries in autocommit-mode.
Slow query log only now.
Print out a line for every query. Show statistics such as the query execution time, error code, number of rows affected and matched.
This options both available as command-line options as config options.
Accept either a filename on the command-line, -, or STDIN, just like normal Unix applications.
count of threads for execute read queries.
Later - not now.
This option notice, what tool should run “use name;” after establish every connection to mysql server.
None other than the one-line-per-query output.
The “options” is mysql settings (such as user, socket, host, port, etc) what should propagate to establishment connection for every mysql's connection.
We'll write awk/perl/sed scripts against the raw output and figure out what is useful.
Result should be single static-linked binaries what run on every system (economy of time).
We should prepare some performance what investigate replay log speed. Also we can collect some internal statistic about load of different components.
We use CMake because is wonderful for C/C++ project (simpler that autotools, out-of-source build, Makefile generator, flexible depends management, support may compilers and libraries).
Single thread to parse input slow query log, (CORE_NUMBER - 2) threads for read queries, 1 thread for write (UPDATE/INSERT) queries. All queries run as fast as possible.