Recovered executor num in yarn-client mode

[yuemeng1]

Hi, all
  I have (maybe a clumsy) question about executor recovery num in 
yarn-client mode. My situation is as follows:

  We have a 1(resource manager) + 3(node manager) cluster, a app is 
running with one driver on the resource manager and 12 executors on all 
the node managers,
and there are 4 executors on each node manager machine. For some reason 
4 executors on one machine disassociated/failed, then 2 executors recovered.
  My question is why not 4 executors recovered? who and how decide the 
number of recovered executors?

 Thanks