K-Means Clustering in Py-spark

I have a really big data frame. I am considering 8 columns to make 8 clusters. Now, normally the code would discard a row even if 1 of the 8 column values is NaN. In my case, I have to write a pyspark code that behaves in such a way that(Even When a row has some NaN values, the code will ignore them and take other column values into consideration and thereby performing clustering even for rows that have some NaN values)Any help or lead is highly appreciated.