HDFS metadata consists of two parts: Base filesystem table (stored in a file called fsimage)
The edit logs which lists changes made to the base table stored in files called edits. Checkpointing is a process of reconciling fsimage with edits to produce a new version of fsimage.
There are two benefits arising out of this: A more recent version of fsimage, and a truncated edit logs. The following properties can help to set how often Checkpointing happens: dfs.namenode.checkpoint.period - Controls how often this reconciliation will be triggered. The number of seconds between two periodic checkpoints; fsimage will be updated and edit log truncated. Checkpiont is not cheap, so there is a balance between running it too often and letting the edit log grow too large. This parameter should be set to get a good balance assuming typical filesystem use in your cluster. dfs.namenode.checkpoint.edits.dir - Determines where on the local filesystem the DFS secondary name node should store the temporary edits to merge. If this is a comma-delimited list of directories then the edits is replicated in all of the directories for redundancy. Default value is same as dfs.namenode.checkpoint.dir dfs.namenode.checkpoint.txns - The Secondary NameNode or CheckpointNode will create a checkpoint of the namespace every 'dfs.namenode.checkpoint.txns' transactions, regardless of whether 'dfs.namenode.checkpoint.period' has expired. dfs.ha.standby.checkpoints - If true, a NameNode in Standby state periodically takes a checkpoint of the namespace, saves it to its local storage and then upload to the remote NameNode.
Also, if you would like to manually checkpoint you can follow: https://community.hortonworks.com/content/supportkb/49438/how-to-manually-checkpoint.html
... View more
