Keras-Tuner always returns the same amount of nodes for my

[MbProg]

I use keras-tuner with RandomSearch to get the number of nodes for my two hidden layers where the second layer must not have more nodes than the first layer and at least 60% of nodes of the first layer. Therefore, I use the following code:

class SimpleMLP(kt.HyperModel):
def build(self,hp):
    first_layer_nodes = hp.Int(name='first_layer_nodes',min_value = 50, max_value=1000,step=32)
    second_layer_nodes = hp.Int(name='second_layer_nodes',min_value = int(first_layer_nodes*0.6), max_value=first_layer_nodes, step = 32)

I get random amount of nodes for first_layer_nodes in the given interval. But for second_layer_nodes I always get 30. And I get 30 even first_layer_nodes is 300 or 900 or anything else. It is out of the given interval of min_value and max_value.
What is wrong with the code?