Refresh HDFS Datanodes after adding Data Optimizer volumes

After you have configured HDFS data nodes to use Data Optimizer volumes, refresh or restart the data nodes so they can register the configuration change and use the volumes. The data node property that was modified is a refreshable configuration, so the data nodes can pick up new data directories without restarting, which can be disruptive.

Typically, Cloudera Manager prompts you to refresh your data nodes because Cloudera Manager detects that the configuration change is refreshable. Click the refresh icon when it appears at the top of the HDFS configuration page to refresh your data nodes. After the refresh, the data nodes can start tiering to Data Optimizer.

In some cases, such as if you created a new configuration role group, Cloudera Manager may prompt for a restart. In this case, contact the Data Optimizer implementation team and your Cluster Administrator to determine how best to proceed and resolve the required restart notifications. You still might be able to perform a refresh by executing the Refresh Cluster action. After the refresh, data nodes are ready to begin tiering as described in the previous paragraph.

CAUTION:

After HDFS has been configured to use Data Optimizer Volumes, do not stop the Volume on a node when the Datanode service is still running. Doing so can result in unpredictable behavior, including HDFS marking a volume as failed.

To verify Data Optimizer is tiering properly, see Tiering HDFS Blocks to Data Optimizer.