Why are the HDFS edit logs retained after a checkpoint operation is complete?

[#1]

In an HA environment, I see a fairly current fsimage file in the Namenode directory. But I also see edit files for the last 20+ days even though the checkpoint frequency is 6 hours/1M transactions.

[#2]

@terry

See this http://hortonworks.com/blog/hdfs-metadata-directories-explained/

[#3]

Pointed me to the right place to do more research.

[#4 · Accepted solution]

The hdfs-site.xml property dfs.namenode.num.extra.edits.retained defaults to a value of 1000000 (1 million) Reducing this value will lower the number of edit logs retained after a checkpoint in complete.