conn_a = rdd.filter(
        lambda r: r[1]['header']['op'] == 7
    ).map(
        lambda r: r[1]
    ).collect()
    conn_d = {str(k['header']['topic']):k for k in conn_a}

# see topic names
    conn_d.keys()e

def topic( a:org.apache.hadoop.io.MapWritable) : Boolean = {
  // return true if we have a header/topic