Implementing CTE on HDFS on a Single Host

It is possible, though not recommended, that an HDFS NameNode and DataNode exist as separate processes on the same host. If this is your deployment, use the following CTE deployment guidelines:

1. Configure the HDFS NameNodes (see Configure NameNodes):

   The directories specified by dfs.datanode.data.dir already exist on the local file system so you do not have to create them.

2. Pass the following information to the Administrator:

   • The dfs.datanode.data.dir directory list and instructions to create a GuardPoint for the HDFS Host Group on each of these directories.

   • Instructions to add the NameNodes IP addresses, or host names, to the HDFS Host Group.

3. Create an HDFS host/client group and host/client group GuardPoint:

   1. Administrator must create an HDFS Host Group to contain the HDFS nodes.

   2. The Administrator must create a GuardPoint for the Host Group on each of the directories specified by dfs.datanode.data.dir

4. Take the DataNode offline and perform a data transformation.

5. Add the NameNode/DataNode host/client to the host/client group.