HDFS dependency on running Ozone
Although Ozone is an independent service and does not use any runtime component of HDFS, Ozone relies on the mechanism of how Cloudera Manager distributes client configurations to different services such as Hive, Spark, YARN, and so on. Because this mechanism of Cloudera Manager does not work without HDFS, Ozone requires HDFS as a dependency when running on a CDP cluster.
If applications such as Hive, Spark, YARN, or others require the cluster HDFS client configurations, Ozone client configurations are also bundled along with the HDFS configurations. Therefore, out-of-the-box support for Ozone based storage can be made available in Data Warehousing and Machine Learning applications. This dependency also implies that changing any Ozone client configuration requires restarting of all the dependent client services.