How to limit Spark concurrent tasks number in one job?

Hi All,
it is possible make fair scheduler pools pluggable? so that we can
implement our own SchedulingAlgorithm. In our case, we want to limit the
max tasks number of one job which will load data from mysql database, if we
set a bigger executer.number * cores.number, it will trigger alarm. Or we
can do this in an other way?