A Checkpoint node in HDFS periodically fetches fsimage and edits from NameNode, and merges them. This merge result is called a Checkpoint.
Once a Checkpoint is created, Checkpoint Node uploads the Checkpoint to NameNode.
Secondary node also takes Checkpoint similar to Checkpoint Node. But it does not upload the Checkpoint to
NameNode.
Main benefit of Checkpoint Node is in case of any failure on NameNode. A NameNode does not merge its edits to fsimage automatically during the runtime. If we have long running task, the edits will become huge. When we restart NameNode, it will take much longer time, because it will first merge the edits. In such a scenario, Checkpoint node helps for a long running task.
Checkpoint nodes performs the task of merging the edits with fsimage and then uploads these to NameNode. This saves time during the restart of NameNode.