The modified way is the below:

df2.cache() # Marks the df2 as cachable
df2.count()  # Action to trigger computation and caching
df2.show()