The secondary Namenode constantly reads the data from the RAM of the Namenode and writes it into the hard disk or the file system. It is not a substitute to the Namenode, so if the Namenode fails, the entire Hadoop system goes down.