A NameNode checkpoint merges the fsimage and edit logs into a newer namespace image. It reduces edit-log replay time and creates a recent metadata image that can support recovery planning.

In non-HA clusters, the SecondaryNameNode usually performs checkpoints. In HA clusters, checkpoint behavior belongs to standby NameNodes, so do not apply the same operational assumptions blindly.

A checkpoint is metadata protection, not a complete data backup. HDFS block replicas still live on DataNodes and need separate cluster backup or replication planning.

Steps to create a Hadoop NameNode checkpoint:

  1. Confirm the checkpoint role configured for the cluster. 
    $ hdfs getconf -secondaryNameNodes
secondary01.example.net
  2. Check HDFS health before forcing a namespace save. 
    $ hdfs fsck /
Status: HEALTHY
 Total blocks (validated): 42

    Related: How to check HDFS cluster health

  3. Enter safe mode before a manual namespace save when required by the operating procedure. 
    $ hdfs dfsadmin -safemode enter
Safe mode is ON

    Related: How to manage HDFS safe mode

  4. Save the namespace. 
    $ hdfs dfsadmin -saveNamespace
Save namespace successful
  5. Leave safe mode after the save completes. 
    $ hdfs dfsadmin -safemode leave
Safe mode is OFF
  6. Verify a recent fsimage exists in the NameNode directory. 
    $ ls -lh /data/hadoop/hdfs/name/current
-rw-r--r-- 1 hdfs hadoop  24M Jun 17 03:42 fsimage_0000000000000012842
-rw-r--r-- 1 hdfs hadoop   62 Jun 17 03:42 fsimage_0000000000000012842.md5