ParameterServer and Worker blocks even though coordinator starts to schedules tasks

Question

I am implementing a simple distributed training app using ParameterServerStrategy from Tensorflow v2.4.1 by following official instructions from tf.distribute.experimental.Param