Relocate the default file system from HDFS to an ECS bucket
Although the system is now usable and may appear to work well, a configuration with HDFS as the default file system is not supported. You must therefore relocate the default file system from HDFS to the root ECS bucket. This procedure copies all files from the HDFS file system to an ECS bucket and then sets the ECS bucket as the default file system.
Use Ambari to stop all services except HDFS, YARN, and Zookeeper.
Copy all existing files on the DAS HDFS file system to the ECS bucket. Even for a new installation of Hadoop, there are critical directories that must exist in the default Hadoop file system. Use
DistCp to perform the file copy.
Table 1. Hadoop configuration to enable Hive concurrency and ACID transactionsThe table lists the Hadoop configuration details to enabe Hive concurrency and ACID transactions
Hadoop location
Property
Value (example)
HDFS Advanced core-site
fs.defaultFS
viprfs://<bucket_name>.<namespace>.<federation_name> For example:
viprfs://mycluster1-root.ns1.federation1
Spark Advanced spark-defaults
spark.eventLog.dir
viprfs://<bucket_name>.<namespace>.<federation>/<spark-history> For example:
viprfs://mycluster1-root.ns1.federation1/spark-history
Spark Advanced spark-defaults
spark.history.fs.logDirectory
viprfs://<bucket_name>.<namespace>.<federation>/<spark-history> For example:
viprfs://mycluster1-root.ns1.federation1/spark-history
Use Ambari to stop and start all services.
Ensure proper directory permissions. If
DistCp encounters any errors, the necessary permissions may not have been applied to critical directories. The following commands set the correct permissions.