Moving Secondary NameNode Checkpoint directory

[#1]

Hi,

The Secondary NameNode (SNN) checkpoint directory at /tmp/hadoop/hdfs/namesecondary is filling up tmp and I would like to move it. What is the best way to do this? Also once moved, can I just RM -R all data in the old directory or do I need to copy it over?

Thanks,

[#2 · Accepted solution]

From Ambari perspective, if you stop HDFS, modify the config property shown below, move all content to the new location it will likely work.

Someone from HDFS team can confirm if there is any affinity to the actual folder path.

<property>
      <name>dfs.namenode.checkpoint.dir</name>
      <value>/hadoop/hdfs/namesecondary</value>
</property>

[#3]

HDFS has no affinity to a specific path for this setting.

[#4]

Apart from the fact that the partition is getting full, the main reason I see to move the checkpoint directory is that you cannot trust that the data under /tmp won't be wiped out after a reboot of your server.

In general, avoid putting any kind of Hadoop information (data or metadata) under /tmp, unless you are sure this is kind of temporary or non critical information.